├── .gitignore ├── nodes ├── modify_hy_model_node.py ├── hy_feta_enhance_node.py ├── hy_attn_override_node.py ├── hy_model_pred_nodes.py ├── rectified_sampler_nodes.py ├── hy_regional_cond_nodes.py ├── flowedit_nodes.py ├── flow_edit_nodes.py └── wrapper_flow_edit_nodes.py ├── utils ├── mask_utils.py ├── feta_enhance_utils.py ├── latent_preview.py └── rope_utils.py ├── __init__.py ├── README.md ├── modules ├── hy_layers.py └── hy_model.py ├── example_workflows ├── example_hy_flowedit.json ├── wrapper_example_hunyuan_flowedit.json └── example_hy_regional_prompting_t2v.json └── LICENSE /.gitignore: -------------------------------------------------------------------------------- 1 | __pycache__ 2 | .DS_Store 3 | -------------------------------------------------------------------------------- /nodes/modify_hy_model_node.py: -------------------------------------------------------------------------------- 1 | from ..modules.hy_layers import inject_blocks 2 | from ..modules.hy_model import inject_model 3 | 4 | 5 | class ConfigureModifiedHYNode: 6 | @classmethod 7 | def INPUT_TYPES(s): 8 | return {"required": { 9 | "model": ("MODEL",), 10 | }} 11 | RETURN_TYPES = ("MODEL",) 12 | 13 | CATEGORY = "hunyuanloom" 14 | FUNCTION = "apply" 15 | 16 | def apply(self, model): 17 | inject_model(model.model.diffusion_model) 18 | inject_blocks(model.model.diffusion_model) 19 | return (model,) 20 | 21 | -------------------------------------------------------------------------------- /nodes/hy_feta_enhance_node.py: -------------------------------------------------------------------------------- 1 | DEFAULT_ATTN = { 2 | 'double': [i for i in range(0, 100, 1)],#[0,1,2,3,4,5,6,7,9,11,13,15,17,19,21,23,25], 3 | 'single': [i for i in range(0, 100, 1)] 4 | } 5 | 6 | class HYFetaEnhanceNode: 7 | @classmethod 8 | def INPUT_TYPES(s): 9 | return {"required": { 10 | "model": ("MODEL",), 11 | "feta_weight": ("FLOAT", {"default": 2, "min": -100.0, "max": 100.0, "step":0.01}), 12 | }, "optional": { 13 | "attn_override": ("ATTN_OVERRIDE",) 14 | }} 15 | RETURN_TYPES = ("MODEL",) 16 | 17 | CATEGORY = "hunyuanloom" 18 | FUNCTION = "apply" 19 | 20 | def apply(self, model, feta_weight, attn_override=DEFAULT_ATTN): 21 | model = model.clone() 22 | 23 | model_options = model.model_options.copy() 24 | transformer_options = model_options['transformer_options'].copy() 25 | 26 | transformer_options['feta_weight'] = feta_weight 27 | transformer_options['feta_layers'] = attn_override 28 | model_options['transformer_options'] = transformer_options 29 | 30 | model.model_options = model_options 31 | return (model,) 32 | 33 | -------------------------------------------------------------------------------- /nodes/hy_attn_override_node.py: -------------------------------------------------------------------------------- 1 | 2 | def is_integer(string): 3 | try: 4 | int(string) 5 | return True 6 | except ValueError: 7 | return False 8 | 9 | 10 | class HYAttnOverrideNode: 11 | @classmethod 12 | def INPUT_TYPES(s): 13 | return { 14 | "required": { 15 | "double_blocks": ("STRING", { "multiline": True }), 16 | "single_blocks": ("STRING", { "multiline": True }), 17 | } 18 | } 19 | 20 | RETURN_TYPES = ("ATTN_OVERRIDE",) 21 | FUNCTION = "build" 22 | 23 | CATEGORY = "hunyuanloom/attn" 24 | 25 | def build(self, double_blocks, single_blocks): 26 | 27 | double_block_map = set() 28 | for block in double_blocks.split(','): 29 | block = block.strip() 30 | if is_integer(block): 31 | double_block_map.add(int(block)) 32 | 33 | single_block_map = set() 34 | for block in single_blocks.split(','): 35 | block = block.strip() 36 | if is_integer(block): 37 | single_block_map.add(int(block)) 38 | 39 | return ({ "double": double_block_map, "single": single_block_map },) 40 | 41 | -------------------------------------------------------------------------------- /utils/mask_utils.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | def consolidate_masks(masks: torch.Tensor, num_latents: int, method: str) -> torch.Tensor: 4 | # masks are floats (e.g., 0.0 or 1.0), do not convert to boolean 5 | 6 | n_frames = masks.shape[0] 7 | frames_per_latent = 4 8 | expected_frames = (num_latents * frames_per_latent) - 3 9 | 10 | if n_frames == expected_frames: 11 | # Perfect match 12 | pass 13 | elif n_frames == 1 or method == "first_only": 14 | # If only one frame, repeat it for all latents 15 | return masks[:1].repeat(num_latents, 1, 1) 16 | elif n_frames > expected_frames: 17 | # Truncate if there are more frames than expected 18 | masks = masks[:expected_frames] 19 | n_frames = expected_frames 20 | else: 21 | # Not enough frames 22 | raise ValueError( 23 | f"For {num_latents} latents, expected {expected_frames} frames " 24 | f"((num_latents * 4) - 3), but got {n_frames} frames" 25 | ) 26 | 27 | # Add padding frames to form a multiple of frames_per_latent 28 | padding_frame = masks[-1:].clone() 29 | padded_masks = torch.cat([masks, padding_frame.repeat(3, 1, 1)], dim=0) 30 | 31 | # Reshape into (num_latents, frames_per_latent, H, W) 32 | grouped_masks = padded_masks.reshape(num_latents, frames_per_latent, *masks.shape[1:]) 33 | 34 | if method == "select_first": 35 | return grouped_masks[:, 0] 36 | elif method == "select_last": 37 | return grouped_masks[:, -1] 38 | elif method == "union": 39 | return 1 - torch.clamp(grouped_masks.sum(dim=1), max=1) 40 | else: 41 | raise ValueError(f"Unknown consolidation method: {method}") 42 | -------------------------------------------------------------------------------- /utils/feta_enhance_utils.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from einops import rearrange 3 | 4 | 5 | def _feta_score(query_image, key_image, head_dim, num_frames, enhance_weight): 6 | scale = head_dim**-0.5 7 | query_image = query_image * scale 8 | attn_temp = query_image @ key_image.transpose(-2, -1) # translate attn to float32 9 | attn_temp = attn_temp.to(torch.float32) 10 | attn_temp = attn_temp.softmax(dim=-1) 11 | 12 | # Reshape to [batch_size * num_tokens, num_frames, num_frames] 13 | attn_temp = attn_temp.reshape(-1, num_frames, num_frames) 14 | 15 | # Create a mask for diagonal elements 16 | diag_mask = torch.eye(num_frames, device=attn_temp.device).bool() 17 | diag_mask = diag_mask.unsqueeze(0).expand(attn_temp.shape[0], -1, -1) 18 | 19 | # Zero out diagonal elements 20 | attn_wo_diag = attn_temp.masked_fill(diag_mask, 0) 21 | 22 | # Calculate mean for each token's attention matrix 23 | # Number of off-diagonal elements per matrix is n*n - n 24 | num_off_diag = num_frames * num_frames - num_frames 25 | mean_scores = attn_wo_diag.sum(dim=(1, 2)) / num_off_diag 26 | 27 | enhance_scores = mean_scores.mean() * (num_frames + enhance_weight) 28 | enhance_scores = enhance_scores.clamp(min=1) 29 | return enhance_scores 30 | 31 | 32 | @torch.compiler.disable() 33 | def get_feta_scores(img_q, img_k, transformer_options): 34 | num_frames = transformer_options['original_shape'][2] 35 | _, num_heads, ST, head_dim = img_q.shape 36 | spatial_dim = ST // num_frames 37 | 38 | query_image = rearrange( 39 | img_q, "B N (T S) C -> (B S) N T C", T=num_frames, S=spatial_dim, N=num_heads, C=head_dim 40 | ) 41 | key_image = rearrange(img_k, "B N (T S) C -> (B S) N T C", T=num_frames, S=spatial_dim, N=num_heads, C=head_dim) 42 | weight = transformer_options.get('feta_weight', 0) 43 | return _feta_score(query_image, key_image, head_dim, num_frames, weight) -------------------------------------------------------------------------------- /__init__.py: -------------------------------------------------------------------------------- 1 | from .nodes.modify_hy_model_node import ConfigureModifiedHYNode 2 | from .nodes.hy_model_pred_nodes import HYInverseModelSamplingPredNode, HYReverseModelSamplingPredNode 3 | from .nodes.rectified_sampler_nodes import HYForwardODESamplerNode, HYReverseODESamplerNode 4 | from .nodes.flowedit_nodes import HYFlowEditGuiderNode, HYFlowEditGuiderAdvNode, HYFlowEditSamplerNode, HYFlowEditGuiderCFGNode, HYFlowEditGuiderCFGAdvNode 5 | 6 | from .nodes.hy_regional_cond_nodes import HYApplyRegionalCondsNode, HYCreateRegionalCondNode 7 | from .nodes.hy_attn_override_node import HYAttnOverrideNode 8 | 9 | from .nodes.hy_feta_enhance_node import HYFetaEnhanceNode 10 | 11 | from .nodes.wrapper_flow_edit_nodes import HyVideoFlowEditSamplerNode 12 | 13 | 14 | NODE_CLASS_MAPPINGS = { 15 | "ConfigureModifiedHY": ConfigureModifiedHYNode, 16 | # RF-Inversion 17 | "HYInverseModelSamplingPred": HYInverseModelSamplingPredNode, 18 | "HYReverseModelSamplingPred": HYReverseModelSamplingPredNode, 19 | "HYForwardODESampler": HYForwardODESamplerNode, 20 | "HYReverseODESampler": HYReverseODESamplerNode, 21 | # FlowEdit 22 | "HYFlowEditGuider": HYFlowEditGuiderNode, 23 | "HYFlowEditGuiderAdv": HYFlowEditGuiderAdvNode, 24 | "HYFlowEditGuiderCFG": HYFlowEditGuiderCFGNode, 25 | "HYFlowEditGuiderCFGAdv": HYFlowEditGuiderCFGAdvNode, 26 | "HYFlowEditSampler": HYFlowEditSamplerNode, 27 | # Regional 28 | "HYApplyRegionalConds": HYApplyRegionalCondsNode, 29 | "HYCreateRegionalCond": HYCreateRegionalCondNode, 30 | "HYAttnOverride": HYAttnOverrideNode, 31 | # Enhance 32 | "HYFetaEnhance": HYFetaEnhanceNode, 33 | # Wrapper 34 | "HyVideoFlowEditSamplerWrapper": HyVideoFlowEditSamplerNode, 35 | } 36 | 37 | NODE_DISPLAY_NAME_MAPPINGS = { 38 | "ConfigureModifiedHY": "Modify Hunyuan Model", 39 | # RF-Inversion 40 | "HYInverseModelSamplingPred": "HY Inverse Model Pred", 41 | "HYReverseModelSamplingPred": "HY Reverse Model Pred", 42 | "HYForwardODESampler": "HY RF-Inv Forward Sampler", 43 | "HYReverseODESampler": "HY RF-Inv Reverse Sampler", 44 | # FlowEdit 45 | "HYFlowEditGuider": "HY FlowEdit Guider", 46 | "HYFlowEditGuiderCFG": "HY FlowEdit Guider CFG", 47 | "HYFlowEditGuiderCFGAdv": "HY FlowEdit Guider CFG Adv.", 48 | "HYFlowEditSampler": "HY FlowEdit Sampler", 49 | # Regional 50 | "HYApplyRegionalConds": "HY Apply Regional Conds", 51 | "HYCreateRegionalCond": "HY Create Regional Cond", 52 | "HYAttnOverride": "HY Attention Override", 53 | # Enhance 54 | "HYFetaEnhance": "HY Feta Enhance", 55 | # Wrapper 56 | "HyVideoFlowEditSamplerWrapper": "HunyuanVideo Flow Edit Sampler (Wrapper)", 57 | } 58 | -------------------------------------------------------------------------------- /nodes/hy_model_pred_nodes.py: -------------------------------------------------------------------------------- 1 | import comfy.sd 2 | import comfy.model_sampling 3 | import comfy.latent_formats 4 | import nodes 5 | 6 | 7 | class InverseCONST: 8 | def calculate_input(self, sigma, noise): 9 | return noise 10 | 11 | def calculate_denoised(self, sigma, model_output, model_input): 12 | sigma = sigma.view(sigma.shape[:1] + (1,) * (model_output.ndim - 1)) 13 | return model_output 14 | 15 | def noise_scaling(self, sigma, noise, latent_image, max_denoise=False): 16 | return latent_image 17 | 18 | def inverse_noise_scaling(self, sigma, latent): 19 | return latent 20 | 21 | 22 | class HYInverseModelSamplingPredNode: 23 | @classmethod 24 | def INPUT_TYPES(s): 25 | return {"required": { "model": ("MODEL",), 26 | "shift": ("FLOAT", {"default": 7, "min": 0.0, "max": 100.0, "step":0.01}), 27 | }} 28 | 29 | RETURN_TYPES = ("MODEL",) 30 | FUNCTION = "patch" 31 | 32 | CATEGORY = "hunyuanloom" 33 | 34 | def patch(self, model, shift): 35 | m = model.clone() 36 | 37 | sampling_base = comfy.model_sampling.ModelSamplingDiscreteFlow 38 | sampling_type = InverseCONST 39 | 40 | class ModelSamplingAdvanced(sampling_base, sampling_type): 41 | pass 42 | 43 | model_sampling = ModelSamplingAdvanced(model.model.model_config) 44 | model_sampling.set_parameters(shift=shift, multiplier=1000) 45 | m.add_object_patch("model_sampling", model_sampling) 46 | return (m, ) 47 | 48 | 49 | class ReverseCONST: 50 | def calculate_input(self, sigma, noise): 51 | return noise 52 | 53 | def calculate_denoised(self, sigma, model_output, model_input): 54 | return model_output # model_input - model_output * sigma 55 | 56 | def noise_scaling(self, sigma, noise, latent_image, max_denoise=False): 57 | return latent_image 58 | 59 | def inverse_noise_scaling(self, sigma, latent): 60 | return latent / (1.0 - sigma) 61 | 62 | 63 | class HYReverseModelSamplingPredNode: 64 | @classmethod 65 | def INPUT_TYPES(s): 66 | return {"required": { "model": ("MODEL",), 67 | "shift": ("FLOAT", {"default": 7, "min": 0.0, "max": 100.0, "step":0.01}), 68 | }} 69 | 70 | RETURN_TYPES = ("MODEL",) 71 | FUNCTION = "patch" 72 | 73 | CATEGORY = "hunyuanloom" 74 | 75 | def patch(self, model, shift): 76 | m = model.clone() 77 | 78 | sampling_base = comfy.model_sampling.ModelSamplingDiscreteFlow 79 | sampling_type = ReverseCONST 80 | 81 | class ModelSamplingAdvanced(sampling_base, sampling_type): 82 | pass 83 | 84 | model_sampling = ModelSamplingAdvanced(model.model.model_config) 85 | model_sampling.set_parameters(shift=shift, multiplier=1000) 86 | m.add_object_patch("model_sampling", model_sampling) 87 | return (m, ) 88 | -------------------------------------------------------------------------------- /utils/latent_preview.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from PIL import Image 3 | 4 | from comfy.cli_args import args, LatentPreviewMethod 5 | import comfy.model_management 6 | import comfy.utils 7 | 8 | MAX_PREVIEW_RESOLUTION = args.preview_size 9 | 10 | def preview_to_image(latent_image): 11 | latents_ubyte = (((latent_image + 1.0) / 2.0).clamp(0, 1) # change scale from -1..1 to 0..1 12 | .mul(0xFF) # to 0..255 13 | ).to(device="cpu", dtype=torch.uint8, non_blocking=comfy.model_management.device_supports_non_blocking(latent_image.device)) 14 | 15 | return Image.fromarray(latents_ubyte.numpy()) 16 | 17 | class LatentPreviewer: 18 | def decode_latent_to_preview(self, x0): 19 | pass 20 | 21 | def decode_latent_to_preview_image(self, preview_format, x0): 22 | preview_image = self.decode_latent_to_preview(x0) 23 | return ("GIF", preview_image, MAX_PREVIEW_RESOLUTION) 24 | 25 | class Latent2RGBPreviewer(LatentPreviewer): 26 | def __init__(self): 27 | latent_rgb_factors = [[-0.41, -0.25, -0.26], 28 | [-0.26, -0.49, -0.24], 29 | [-0.37, -0.54, -0.3], 30 | [-0.04, -0.29, -0.29], 31 | [-0.52, -0.59, -0.39], 32 | [-0.56, -0.6, -0.02], 33 | [-0.53, -0.06, -0.48], 34 | [-0.51, -0.28, -0.18], 35 | [-0.59, -0.1, -0.33], 36 | [-0.56, -0.54, -0.41], 37 | [-0.61, -0.19, -0.5], 38 | [-0.05, -0.25, -0.17], 39 | [-0.23, -0.04, -0.22], 40 | [-0.51, -0.56, -0.43], 41 | [-0.13, -0.4, -0.05], 42 | [-0.01, -0.01, -0.48]] 43 | self.latent_rgb_factors = torch.tensor(latent_rgb_factors, device="cpu").transpose(0, 1) 44 | self.latent_rgb_factors_bias = torch.tensor([0.138, 0.025, -0.299], device="cpu") 45 | 46 | def decode_latent_to_preview(self, x0): 47 | self.latent_rgb_factors = self.latent_rgb_factors.to(dtype=x0.dtype, device=x0.device) 48 | if self.latent_rgb_factors_bias is not None: 49 | self.latent_rgb_factors_bias = self.latent_rgb_factors_bias.to(dtype=x0.dtype, device=x0.device) 50 | 51 | latent_image = torch.nn.functional.linear(x0[0].permute(1, 2, 0), self.latent_rgb_factors, 52 | bias=self.latent_rgb_factors_bias) 53 | return preview_to_image(latent_image) 54 | 55 | 56 | def get_previewer(): 57 | previewer = None 58 | method = args.preview_method 59 | if method != LatentPreviewMethod.NoPreviews: 60 | # TODO previewer method 61 | 62 | if method == LatentPreviewMethod.Auto: 63 | method = LatentPreviewMethod.Latent2RGB 64 | 65 | if previewer is None: 66 | previewer = Latent2RGBPreviewer() 67 | return previewer 68 | 69 | def prepare_callback(model, steps, x0_output_dict=None): 70 | preview_format = "JPEG" 71 | if preview_format not in ["JPEG", "PNG"]: 72 | preview_format = "JPEG" 73 | 74 | previewer = get_previewer() 75 | 76 | pbar = comfy.utils.ProgressBar(steps) 77 | def callback(step, x0, x, total_steps): 78 | if x0_output_dict is not None: 79 | x0_output_dict["x0"] = x0 80 | preview_bytes = None 81 | if previewer: 82 | preview_bytes = previewer.decode_latent_to_preview_image(preview_format, x0) 83 | pbar.update_absolute(step + 1, total_steps, preview_bytes) 84 | return callback 85 | 86 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # ComfyUI-HunyuanLoom (WIP) 2 | A set of nodes to edit videos using the Hunyuan Video model 3 | 4 | ## Installation 5 | These nodes are for the Hunyuan Video model. If you haven't installed it yet, you can follow [these examples](https://comfyanonymous.github.io/ComfyUI_examples/hunyuan_video/). 6 | 7 | 8 | ## FlowEdit 9 | 10 | This is an implementation of [FlowEdit](https://github.com/fallenshock/FlowEdit). 11 | 12 | Use [this workflow](https://github.com/logtd/ComfyUI-HunyuanLoom/blob/main/example_workflows/example_hy_flowedit.json) to get started. 13 | 14 | https://github.com/user-attachments/assets/91e38df0-1725-4a6f-9d7d-214baa0758ab 15 | 16 | ### Settings Overview 17 | flowedit_settings 18 | 19 | Note I took some liberties in naming these settings as the original research called them `n_min`, `n_max`, etc. 20 | There are three kinds of steps in FlowEdit: 21 | 1. skip steps -- these come at the beginning and are literally skipping steps 22 | 2. diff steps -- in these settings they are unnamed and are the remaining steps from `total - skip_steps - drift_steps` 23 | 3. drift steps -- these steps come at the end. They can essentially be thought of as normal steps that you'd use in sampling. They are named `drift` because there's nothing controlling them from drifting away from the input video. 24 | 25 | #### Skip Steps 26 | The more steps you skip, the less overall power sampling has from changing the source input. Increasing these will significantly impact how much can be changed. 27 | 28 | #### Diff Steps (the middle unnamed steps) 29 | These steps do a comparison between the source prompt and target prompt and attempt to only take the smallest difference needed to make the input video go to the target video. Admittedly, these have very little affect on Hunyuan compared to Flux and LTX, but are still needed. 30 | 31 | When using these steps in Hunyuan not a lot changes, but it does help the sampling move towards the target without going too far from the source. These steps help seed the direction of the generation so that it incorporates the input and the target prompt change. 32 | 33 | #### Drift Steps 34 | These steps come at the end and are essentially normal steps in a sampler. They help remove blur that can happen and "refine" the video. Also in Hunyuan cause the biggest changes to the input video. 35 | 36 | ## Regional Prompting (Experimental) 37 | 38 | > [!IMPORTANT] 39 | > These nodes are experimental in their current state. They require some tuning to make work right and you should not expect quality results on your first try. 40 | 41 | Regional prompting allows you to prompt specific areas of the video over time. Due to this, you can also give different prompts over time (more nodes to make this easier coming). 42 | The root implementation is based on [InstantX's Regional Prompting for Flux](https://github.com/instantX-research/Regional-Prompting-FLUX/tree/main). 43 | 44 | Use [this workflow](https://github.com/logtd/ComfyUI-HunyuanLoom/blob/main/example_workflows/example_hy_regional_prompting_t2v.json) to get started. 45 | 46 | Example using regional prompts on left and right sides. 47 | 48 | https://github.com/user-attachments/assets/37e34c3e-b85b-416a-a0a6-107a238a1783 49 | 50 | Example using regional prompts for first half of frames and second half of frames. 51 | 52 | https://github.com/user-attachments/assets/f24415b4-5c99-4bef-a5bc-7589b5f8a606 53 | 54 | 55 | ## Acknowledgements 56 | 57 | FlowEdit 58 | ``` 59 | @article{kulikov2024flowedit, 60 | title = {FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models}, 61 | author = {Kulikov, Vladimir and Kleiner, Matan and Huberman-Spiegelglas, Inbar and Michaeli, Tomer}, 62 | journal = {arXiv preprint arXiv:2412.08629}, 63 | year = {2024} 64 | } 65 | ``` 66 | 67 | Regional Prompting 68 | ``` 69 | @article{chen2024training, 70 | title={Training-free Regional Prompting for Diffusion Transformers}, 71 | author={Chen, Anthony and Xu, Jianjin and Zheng, Wenzhao and Dai, Gaole and Wang, Yida and Zhang, Renrui and Wang, Haofan and Zhang, Shanghang}, 72 | journal={arXiv preprint arXiv:2411.02395}, 73 | year={2024} 74 | } 75 | ``` 76 | -------------------------------------------------------------------------------- /modules/hy_layers.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from torch import Tensor 3 | 4 | from comfy.ldm.flux.math import apply_rope 5 | from comfy.ldm.flux.layers import SingleStreamBlock, DoubleStreamBlock 6 | from comfy.ldm.modules.attention import optimized_attention 7 | 8 | from ..utils.feta_enhance_utils import get_feta_scores 9 | 10 | 11 | def attention(q: Tensor, k: Tensor, v: Tensor, pe: Tensor, mask=None, skip_rope=False) -> Tensor: 12 | if not skip_rope: 13 | q, k = apply_rope(q, k, pe) 14 | 15 | heads = q.shape[1] 16 | x = optimized_attention(q, k, v, heads, skip_reshape=True, mask=mask) 17 | return x 18 | 19 | 20 | class ModifiedDoubleStreamBlock(DoubleStreamBlock): 21 | def forward(self, img: Tensor, txt: Tensor, vec: Tensor, pe: Tensor, attn_mask=None, transformer_options={}): 22 | img_mod1, img_mod2 = self.img_mod(vec) 23 | txt_mod1, txt_mod2 = self.txt_mod(vec) 24 | 25 | # prepare image for attention 26 | img_modulated = self.img_norm1(img) 27 | img_modulated = (1 + img_mod1.scale) * img_modulated + img_mod1.shift 28 | img_qkv = self.img_attn.qkv(img_modulated) 29 | img_q, img_k, img_v = img_qkv.view(img_qkv.shape[0], img_qkv.shape[1], 3, self.num_heads, -1).permute(2, 0, 3, 1, 4) 30 | img_q, img_k = self.img_attn.norm(img_q, img_k, img_v) 31 | 32 | # prepare txt for attention 33 | txt_modulated = self.txt_norm1(txt) 34 | txt_modulated = (1 + txt_mod1.scale) * txt_modulated + txt_mod1.shift 35 | txt_qkv = self.txt_attn.qkv(txt_modulated) 36 | txt_q, txt_k, txt_v = txt_qkv.view(txt_qkv.shape[0], txt_qkv.shape[1], 3, self.num_heads, -1).permute(2, 0, 3, 1, 4) 37 | txt_q, txt_k = self.txt_attn.norm(txt_q, txt_k, txt_v) 38 | 39 | mask_fn = transformer_options.get('patches_replace', {}).get(f'double', {}).get(('mask_fn', self.idx), None) 40 | if mask_fn is not None: 41 | attn_mask = mask_fn(None, transformer_options, txt.shape[1]) 42 | 43 | skip_rope = False 44 | q = torch.cat((img_q, txt_q), dim=2) 45 | k = torch.cat((img_k, txt_k), dim=2) 46 | feta_scores = None 47 | if transformer_options.get('feta_weight', 0) > 0 and self.idx in transformer_options['feta_layers']['double']: 48 | skip_rope = True 49 | q, k = apply_rope(q, k, pe) 50 | txt_size = transformer_options['txt_size'] 51 | img_q = q[:,:,:-txt_size] 52 | img_k = k[:,:,:-txt_size] 53 | feta_scores = get_feta_scores(img_q, img_k, transformer_options) 54 | 55 | # run actual attention 56 | attn = attention(q, k, torch.cat((img_v, txt_v), dim=2),pe=pe, mask=attn_mask, skip_rope=skip_rope) 57 | 58 | img_attn, txt_attn = attn[:, : img.shape[1]], attn[:, img.shape[1]:] 59 | 60 | if feta_scores is not None: 61 | img_attn *= feta_scores 62 | 63 | # calculate the img bloks 64 | img = img + img_mod1.gate * self.img_attn.proj(img_attn) 65 | img = img + img_mod2.gate * self.img_mlp((1 + img_mod2.scale) * self.img_norm2(img) + img_mod2.shift) 66 | 67 | # calculate the txt bloks 68 | txt += txt_mod1.gate * self.txt_attn.proj(txt_attn) 69 | txt += txt_mod2.gate * self.txt_mlp((1 + txt_mod2.scale) * self.txt_norm2(txt) + txt_mod2.shift) 70 | 71 | if txt.dtype == torch.float16: 72 | txt = torch.nan_to_num(txt, nan=0.0, posinf=65504, neginf=-65504) 73 | 74 | return img, txt 75 | 76 | 77 | class ModifiedSingleStreamBlock(SingleStreamBlock): 78 | def forward(self, x: Tensor, vec: Tensor, pe: Tensor, attn_mask=None, transformer_options={}) -> Tensor: 79 | mod, _ = self.modulation(vec) 80 | x_mod = (1 + mod.scale) * self.pre_norm(x) + mod.shift 81 | qkv, mlp = torch.split(self.linear1(x_mod), [3 * self.hidden_size, self.mlp_hidden_dim], dim=-1) 82 | 83 | q, k, v = qkv.view(qkv.shape[0], qkv.shape[1], 3, self.num_heads, -1).permute(2, 0, 3, 1, 4) 84 | q, k = self.norm(q, k, v) 85 | 86 | mask_fn = transformer_options.get('patches_replace', {}).get(f'single', {}).get(('mask_fn', self.idx), None) 87 | if mask_fn is not None: 88 | attn_mask = mask_fn(q, transformer_options, None) 89 | 90 | skip_rope = False 91 | feta_scores = None 92 | txt_size = transformer_options['txt_size'] 93 | if transformer_options.get('feta_weight', 0) > 0 and self.idx in transformer_options['feta_layers']['single']: 94 | skip_rope = True 95 | q, k = apply_rope(q, k, pe) 96 | img_q = q[:,:,:-txt_size] 97 | img_k = k[:,:,:-txt_size] 98 | feta_scores = get_feta_scores(img_q, img_k, transformer_options) 99 | 100 | # compute attention 101 | attn = attention(q, k, v, pe=pe, mask=attn_mask, skip_rope=skip_rope) 102 | 103 | if feta_scores is not None: 104 | attn[:,:-txt_size] *= feta_scores 105 | 106 | # compute activation in mlp stream, cat again and run second linear layer 107 | output = self.linear2(torch.cat((attn, self.mlp_act(mlp)), 2)) 108 | x += mod.gate * output 109 | if x.dtype == torch.float16: 110 | x = torch.nan_to_num(x, nan=0.0, posinf=65504, neginf=-65504) 111 | 112 | return x 113 | 114 | 115 | 116 | def inject_blocks(diffusion_model): 117 | for i, block in enumerate(diffusion_model.double_blocks): 118 | block.__class__ = ModifiedDoubleStreamBlock 119 | block.idx = i 120 | 121 | for i, block in enumerate(diffusion_model.single_blocks): 122 | block.__class__ = ModifiedSingleStreamBlock 123 | block.idx = i 124 | 125 | return diffusion_model -------------------------------------------------------------------------------- /modules/hy_model.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from einops import repeat 3 | from torch import Tensor 4 | 5 | from comfy.ldm.flux.layers import ( 6 | timestep_embedding 7 | ) 8 | from comfy.ldm.hunyuan_video.model import HunyuanVideo 9 | 10 | 11 | class ModifiedHunyuanVideo(HunyuanVideo): 12 | def forward_orig( 13 | self, 14 | img: Tensor, 15 | img_ids: Tensor, 16 | txt: Tensor, 17 | txt_ids: Tensor, 18 | txt_mask: Tensor, 19 | timesteps: Tensor, 20 | y: Tensor, 21 | guidance: Tensor = None, 22 | control=None, 23 | transformer_options={}, 24 | ) -> Tensor: 25 | patches_replace = transformer_options.get("patches_replace", {}) 26 | 27 | initial_shape = list(img.shape) 28 | transformer_options['original_shape'] = initial_shape 29 | # running on sequences img 30 | img = self.img_in(img) 31 | vec = self.time_in(timestep_embedding(timesteps, 256, time_factor=1.0).to(img.dtype)) 32 | 33 | vec = vec + self.vector_in(y[:, :self.params.vec_in_dim]) 34 | 35 | if self.params.guidance_embed: 36 | if guidance is None: 37 | raise ValueError("Didn't get guidance strength for guidance distilled model.") 38 | vec = vec + self.guidance_in(timestep_embedding(guidance, 256).to(img.dtype)) 39 | 40 | if txt_mask is not None and not torch.is_floating_point(txt_mask): 41 | txt_mask = (txt_mask - 1).to(img.dtype) * torch.finfo(img.dtype).max 42 | 43 | transformer_options['txt_size'] = txt.shape[1] 44 | txt = self.txt_in(txt, timesteps, txt_mask) 45 | 46 | ids = torch.cat((img_ids, txt_ids), dim=1) 47 | pe = self.pe_embedder(ids) 48 | 49 | img_len = img.shape[1] 50 | if txt_mask is not None: 51 | attn_mask_len = img_len + txt.shape[1] 52 | attn_mask = torch.zeros((1, 1, attn_mask_len), dtype=img.dtype, device=img.device) 53 | attn_mask[:, 0, img_len:] = txt_mask 54 | else: 55 | attn_mask = None 56 | 57 | blocks_replace = patches_replace.get("dit", {}) 58 | for i, block in enumerate(self.double_blocks): 59 | if ("double_block", i) in blocks_replace: 60 | def block_wrap(args): 61 | out = {} 62 | out["img"], out["txt"] = block(img=args["img"], txt=args["txt"], vec=args["vec"], pe=args["pe"], attn_mask=args["attention_mask"]) 63 | return out 64 | 65 | out = blocks_replace[("double_block", i)]({"img": img, "txt": txt, "vec": vec, "pe": pe, "attention_mask": attn_mask}, {"original_block": block_wrap}) 66 | txt = out["txt"] 67 | img = out["img"] 68 | else: 69 | img, txt = block(img=img, txt=txt, vec=vec, pe=pe, attn_mask=attn_mask, transformer_options=transformer_options) 70 | 71 | if control is not None: # Controlnet 72 | control_i = control.get("input") 73 | if i < len(control_i): 74 | add = control_i[i] 75 | if add is not None: 76 | img += add 77 | 78 | img = torch.cat((img, txt), 1) 79 | 80 | for i, block in enumerate(self.single_blocks): 81 | if ("single_block", i) in blocks_replace: 82 | def block_wrap(args): 83 | out = {} 84 | out["img"] = block(args["img"], vec=args["vec"], pe=args["pe"], attn_mask=args["attention_mask"]) 85 | return out 86 | 87 | out = blocks_replace[("single_block", i)]({"img": img, "vec": vec, "pe": pe, "attention_mask": attn_mask}, {"original_block": block_wrap}) 88 | img = out["img"] 89 | else: 90 | img = block(img, vec=vec, pe=pe, attn_mask=attn_mask, transformer_options=transformer_options) 91 | 92 | if control is not None: # Controlnet 93 | control_o = control.get("output") 94 | if i < len(control_o): 95 | add = control_o[i] 96 | if add is not None: 97 | img[:, : img_len] += add 98 | 99 | img = img[:, : img_len] 100 | 101 | img = self.final_layer(img, vec) # (N, T, patch_size ** 2 * out_channels) 102 | 103 | shape = initial_shape[-3:] 104 | for i in range(len(shape)): 105 | shape[i] = shape[i] // self.patch_size[i] 106 | img = img.reshape([img.shape[0]] + shape + [self.out_channels] + self.patch_size) 107 | img = img.permute(0, 4, 1, 5, 2, 6, 3, 7) 108 | img = img.reshape(initial_shape[0], self.out_channels, initial_shape[2], initial_shape[3], initial_shape[4]) 109 | return img 110 | 111 | def forward(self, x, timestep, context, y, guidance, attention_mask=None, control=None, transformer_options={}, **kwargs): 112 | bs, c, t, h, w = x.shape 113 | patch_size = self.patch_size 114 | t_len = ((t + (patch_size[0] // 2)) // patch_size[0]) 115 | h_len = ((h + (patch_size[1] // 2)) // patch_size[1]) 116 | w_len = ((w + (patch_size[2] // 2)) // patch_size[2]) 117 | img_ids = torch.zeros((t_len, h_len, w_len, 3), device=x.device, dtype=x.dtype) 118 | img_ids[:, :, :, 0] = img_ids[:, :, :, 0] + torch.linspace(0, t_len - 1, steps=t_len, device=x.device, dtype=x.dtype).reshape(-1, 1, 1) 119 | img_ids[:, :, :, 1] = img_ids[:, :, :, 1] + torch.linspace(0, h_len - 1, steps=h_len, device=x.device, dtype=x.dtype).reshape(1, -1, 1) 120 | img_ids[:, :, :, 2] = img_ids[:, :, :, 2] + torch.linspace(0, w_len - 1, steps=w_len, device=x.device, dtype=x.dtype).reshape(1, 1, -1) 121 | img_ids = repeat(img_ids, "t h w c -> b (t h w) c", b=bs) 122 | 123 | regional_conditioning = transformer_options.get('patches', {}).get('regional_conditioning', None) 124 | if regional_conditioning is not None: 125 | context = regional_conditioning[0](context, transformer_options) 126 | 127 | txt_ids = torch.zeros((bs, context.shape[1], 3), device=x.device, dtype=x.dtype) 128 | out = self.forward_orig(x, img_ids, context, txt_ids, attention_mask, timestep, y, guidance, control, transformer_options) 129 | return out 130 | 131 | 132 | def inject_model(diffusion_model): 133 | diffusion_model.__class__ = ModifiedHunyuanVideo 134 | return diffusion_model 135 | -------------------------------------------------------------------------------- /nodes/rectified_sampler_nodes.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from tqdm import trange 3 | 4 | from comfy.samplers import KSAMPLER 5 | 6 | 7 | def generate_eta_values(steps, start_time, end_time, eta, eta_trend): 8 | eta_values = [0] * steps 9 | 10 | if eta_trend == 'constant': 11 | for i in range(start_time, end_time): 12 | eta_values[i] = eta 13 | elif eta_trend == 'linear_increase': 14 | for i in range(start_time, end_time): 15 | progress = (i - start_time) / (end_time - start_time - 1) 16 | eta_values[i] = eta * progress 17 | elif eta_trend == 'linear_decrease': 18 | for i in range(start_time, end_time): 19 | progress = 1 - (i - start_time) / (end_time - start_time - 1) 20 | eta_values[i] = eta * progress 21 | 22 | return eta_values 23 | 24 | 25 | 26 | def get_sample_forward(gamma, start_step, end_step, gamma_trend, seed): 27 | # Controlled Forward ODE (Algorithm 1) 28 | 29 | @torch.no_grad() 30 | def sample_forward(model, y0, sigmas, extra_args=None, callback=None, disable=None): 31 | generator = torch.Generator() 32 | generator.manual_seed(seed) 33 | extra_args = {} if extra_args is None else extra_args 34 | Y = y0.clone() 35 | y1 = torch.randn(Y.shape, generator=generator).to(y0.device) 36 | N = len(sigmas)-1 37 | s_in = y0.new_ones([y0.shape[0]]) 38 | gamma_values = generate_eta_values(N, start_step, end_step, gamma, gamma_trend) 39 | for i in trange(N, disable=disable): 40 | # t_i = model.inner_model.inner_model.model_sampling.timestep(sigmas[i]) - old code 41 | t_i = sigmas[i] / max(sigmas) # - new code 42 | 43 | # 6. Unconditional Vector field uti(Yti) = u(Yti, ti, Φ(“”); φ) 44 | unconditional_vector_field = model(Y, s_in * sigmas[i], **extra_args) # this implementation takes sigma instead of timestep 45 | 46 | # 7.Conditional Vector field uti(Yti|y1) = (y1−Yti)/1−ti 47 | conditional_vector_field = (y1-Y)/(1-t_i) 48 | 49 | # 8. Controlled Vector field ti(Yti) = uti(Yti) + γ (uti(Yti|y1) − uti(Yti)) 50 | controlled_vector_field = unconditional_vector_field + gamma_values[i] * (conditional_vector_field - unconditional_vector_field) 51 | 52 | # 9. Next state Yti+1 = Yti + ˆuti(Yti) (σ(ti+1) − σ(ti)) 53 | Y = Y + controlled_vector_field * (sigmas[i+1] - sigmas[i]) 54 | 55 | if callback is not None: 56 | callback({'x': Y, 'denoised': Y, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i]}) 57 | 58 | return Y 59 | 60 | return sample_forward 61 | 62 | 63 | 64 | def get_sample_reverse(latent_image, eta, start_time, end_time, eta_trend): 65 | # Controlled Reverse ODE (Algorithm 2) 66 | @torch.no_grad() 67 | def sample_reverse(model, y1, sigmas, extra_args=None, callback=None, disable=None): 68 | extra_args = {} if extra_args is None else extra_args 69 | X = y1.clone() 70 | N = len(sigmas)-1 71 | y0 = latent_image.clone().to(y1.device) 72 | s_in = y0.new_ones([y0.shape[0]]) 73 | eta_values = generate_eta_values(N, start_time, end_time, eta, eta_trend) 74 | for i in trange(N, disable=disable): 75 | # t_i = 1-model.inner_model.inner_model.model_sampling.timestep(sigmas[i]) # TODO: figure out which one to use 76 | t_i = i/N # Empiracally better results 77 | sigma = sigmas[i] 78 | 79 | # 5. Unconditional Vector field uti(Xti) = -u(Xti, 1-ti, Φ(“prompt”); φ) 80 | unconditional_vector_field = -model(X, sigma*s_in, **extra_args) # this implementation takes sigma instead of timestep 81 | 82 | # 6.Conditional Vector field uti(Xti|y0) = (y0−Xti)/(1−ti) 83 | conditional_vector_field = (y0-X)/(1-t_i) 84 | 85 | # 7. Controlled Vector field ti(Yti) = uti(Yti) + γ (uti(Yti|y1) − uti(Yti)) 86 | controlled_vector_field = unconditional_vector_field + eta_values[i] * (conditional_vector_field - unconditional_vector_field) 87 | 88 | # 8. Next state Yti+1 = Yti + ˆuti(Yti) (σ(ti+1) − σ(ti)) 89 | X = X + controlled_vector_field * (sigmas[i] - sigmas[i+1]) 90 | 91 | if callback is not None: 92 | callback({'x': X, 'denoised': X, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i]}) 93 | 94 | return X 95 | 96 | return sample_reverse 97 | 98 | 99 | class HYForwardODESamplerNode: 100 | @classmethod 101 | def INPUT_TYPES(s): 102 | return {"required": { 103 | "gamma": ("FLOAT", {"default": 0.5, "min": 0.0, "max": 100.0, "step": 0.01}), 104 | "start_step": ("INT", {"default": 0, "min": 0, "max": 1000, "step": 1}), 105 | "end_step": ("INT", {"default": 5, "min": 0, "max": 1000, "step": 1}), 106 | "gamma_trend": (['constant', 'linear_increase', 'linear_decrease'],) 107 | }, "optional": { 108 | "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff }), 109 | }} 110 | RETURN_TYPES = ("SAMPLER",) 111 | FUNCTION = "build" 112 | 113 | CATEGORY = "hunyuanloom" 114 | 115 | def build(self, gamma, start_step, end_step, gamma_trend, seed=0): 116 | sampler = KSAMPLER(get_sample_forward(gamma, start_step, end_step, gamma_trend, seed)) 117 | 118 | return (sampler, ) 119 | 120 | 121 | class HYReverseODESamplerNode: 122 | @classmethod 123 | def INPUT_TYPES(s): 124 | return {"required": { 125 | "model": ("MODEL",), 126 | "latent_image": ("LATENT",), 127 | "eta": ("FLOAT", {"default": 0.8, "min": 0.0, "max": 100.0, "step": 0.01}), 128 | "start_step": ("INT", {"default": 0, "min": 0, "max": 1000, "step": 1}), 129 | "end_step": ("INT", {"default": 5, "min": 0, "max": 1000, "step": 1}), 130 | }, "optional": { 131 | "eta_trend": (['constant', 'linear_increase', 'linear_decrease'],) 132 | }} 133 | RETURN_TYPES = ("SAMPLER",) 134 | FUNCTION = "build" 135 | 136 | CATEGORY = "hunyuanloom" 137 | 138 | def build(self, model, latent_image, eta, start_step, end_step, eta_trend='constant'): 139 | process_latent_in = model.get_model_object("process_latent_in") 140 | latent_image = process_latent_in(latent_image['samples']) 141 | sampler = KSAMPLER(get_sample_reverse(latent_image, eta, start_step, end_step, eta_trend)) 142 | 143 | return (sampler, ) 144 | -------------------------------------------------------------------------------- /nodes/hy_regional_cond_nodes.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | import comfy.model_sampling 4 | 5 | from ..utils.mask_utils import consolidate_masks 6 | 7 | 8 | DEFAULT_REGIONAL_ATTN = { 9 | 'double': [i for i in range(1, 100, 1)],#[0,1,2,3,4,5,6,7,9,11,13,15,17,19,21,23,25], 10 | 'single': [i for i in range(1, 100, 2)] 11 | } 12 | 13 | 14 | class RegionalMask(torch.nn.Module): 15 | def __init__(self, mask: torch.Tensor, start_percent: float, end_percent: float) -> None: 16 | super().__init__() 17 | self.register_buffer('mask', mask) 18 | self.start_percent = start_percent 19 | self.end_percent = end_percent 20 | 21 | def __call__(self, q, transformer_options, *args, **kwargs): 22 | if self.start_percent <= 1 - transformer_options['sigmas'][0] < self.end_percent: 23 | return self.mask 24 | 25 | return None 26 | 27 | 28 | class RegionalConditioning(torch.nn.Module): 29 | def __init__(self, region_cond: torch.Tensor, start_percent: float, end_percent: float, main_cond_strength: float, always_included: bool) -> None: 30 | super().__init__() 31 | self.register_buffer('region_cond', region_cond) 32 | self.start_percent = start_percent 33 | self.end_percent = end_percent 34 | self.main_cond_strength = main_cond_strength 35 | self.always_included = always_included 36 | 37 | def __call__(self, context, transformer_options, *args, **kwargs): 38 | if self.start_percent <= 1 - transformer_options['sigmas'][0] < self.end_percent or self.always_included: 39 | return torch.cat([context*self.main_cond_strength, self.region_cond.to(context.dtype)], dim=1) 40 | return context 41 | 42 | 43 | class HYCreateRegionalCondNode: 44 | @classmethod 45 | def INPUT_TYPES(s): 46 | return {"required": { 47 | "cond": ("CONDITIONING",), 48 | "mask": ("MASK",), 49 | "cond_strength": ("FLOAT", {"default": 1, "min": 0.0, "max": 10.0, "step": 0.01, "round": 0.01}), 50 | "mask_consolidation": (["first_only", "select_first", "select_last", "union"], { "tooltip": "How to choose the masking when they are compressed for the latents temporally."}), 51 | }, "optional": { 52 | "prev_regions": ("REGION_COND",), 53 | }} 54 | 55 | RETURN_TYPES = ("REGION_COND",) 56 | FUNCTION = "create" 57 | 58 | CATEGORY = "hunyuanloom" 59 | 60 | def create(self, cond, mask, cond_strength, mask_consolidation, prev_regions=[]): 61 | prev_regions = [*prev_regions] 62 | prev_regions.append({ 63 | 'mask': mask, 64 | 'cond': cond[0][0] * cond_strength, 65 | 'mask_consolidation': mask_consolidation, 66 | }) 67 | 68 | return (prev_regions,) 69 | 70 | 71 | class HYApplyRegionalCondsNode: 72 | @classmethod 73 | def INPUT_TYPES(s): 74 | return {"required": { 75 | "model": ("MODEL",), 76 | "cond": ("CONDITIONING",), 77 | "region_conds": ("REGION_COND",), 78 | "latent": ("LATENT",), 79 | "start_percent": ("FLOAT", {"default": 0, "min": 0.0, "max": 1.0, "step": 0.01, "round": 0.01}), 80 | "end_percent": ("FLOAT", {"default": 0.5, "min": 0.0, "max": 1.0, "step": 0.01, "round": 0.01}), 81 | "main_cond_strength": ("FLOAT", {"default": 1, "min": 0.0, "max": 10.0, "step": 0.01, "round": 0.01}), 82 | "always_included": ("BOOLEAN", { "default": False, "tooltip": "Whether to keep the prompts always but allow them to affect the entire video (unmasked) outside of end/start percents." }), 83 | }, "optional": { 84 | "attn_override": ("ATTN_OVERRIDE",) 85 | }} 86 | 87 | RETURN_TYPES = ("MODEL",) 88 | FUNCTION = "patch" 89 | 90 | CATEGORY = "hunyuanloom" 91 | 92 | def patch(self, model, cond, region_conds, latent, start_percent, end_percent, main_cond_strength, always_included, attn_override=DEFAULT_REGIONAL_ATTN): 93 | model = model.clone() 94 | 95 | latent = latent['samples'] 96 | b, c, f, h, w = latent.shape 97 | h //=2 98 | w //=2 99 | 100 | img_len = h*w*f 101 | 102 | main_cond = cond[0][0] 103 | main_cond_size = main_cond.shape[1] 104 | 105 | regional_conditioning = torch.cat([region_cond['cond'] for region_cond in region_conds], dim=1) 106 | text_len = main_cond_size + regional_conditioning.shape[1] 107 | 108 | regional_mask = torch.zeros((text_len + img_len, text_len + img_len), dtype=torch.bool) 109 | 110 | self_attend_masks = torch.zeros((img_len, img_len), dtype=torch.bool) 111 | union_masks = torch.zeros((img_len, img_len), dtype=torch.bool) 112 | 113 | main_mask = torch.zeros((f, h, w), dtype=torch.float16) 114 | if main_cond_strength == 0: 115 | main_mask = torch.ones((f, h, w), dtype=torch.float16) 116 | 117 | region_conds = [ 118 | { 119 | 'mask': main_mask, 120 | 'cond': torch.ones((1, main_cond_size, 4096), dtype=torch.float16), 121 | 'mask_consolidation': 'first_only', 122 | }, 123 | *region_conds 124 | ] 125 | 126 | current_seq_len = 0 127 | for region_cond_dict in region_conds: 128 | region_cond = region_cond_dict['cond'] 129 | region_mask = region_cond_dict['mask'] 130 | region_mask = consolidate_masks(region_mask, f, region_cond_dict['mask_consolidation']) 131 | region_mask = 1 - region_mask 132 | region_mask = torch.nn.functional.interpolate(region_mask[None, :, :, :], (h, w), mode='nearest-exact').flatten().unsqueeze(1).repeat(1, region_cond.size(1)) 133 | next_seq_len = current_seq_len + region_cond.shape[1] 134 | 135 | # txt attends to itself 136 | regional_mask[current_seq_len:next_seq_len, current_seq_len:next_seq_len] = True 137 | 138 | # txt attends to corresponding regional img 139 | regional_mask[current_seq_len:next_seq_len, text_len:] = region_mask.transpose(-1, -2) 140 | 141 | # regional img attends to corresponding txt 142 | regional_mask[text_len:, current_seq_len:next_seq_len] = region_mask 143 | 144 | # regional img attends to corresponding regional img 145 | img_size_masks = region_mask[:, :1].repeat(1, img_len) 146 | img_size_masks_transpose = img_size_masks.transpose(-1, -2) 147 | self_attend_masks = torch.logical_or(self_attend_masks, 148 | torch.logical_and(img_size_masks, img_size_masks_transpose)) 149 | 150 | # update union 151 | union_masks = torch.logical_or(union_masks, 152 | torch.logical_or(img_size_masks, img_size_masks_transpose)) 153 | 154 | current_seq_len = next_seq_len 155 | 156 | background_masks = torch.logical_not(union_masks) 157 | background_and_self_attend_masks = torch.logical_or(background_masks, self_attend_masks) 158 | regional_mask[text_len:, text_len:] = background_and_self_attend_masks 159 | 160 | # Split rows into text and image 161 | text_rows, img_rows = regional_mask.split([text_len, img_len], dim=0) 162 | 163 | # Now split each part into text and image columns 164 | text_rows_text_cols, text_rows_img_cols = text_rows.split([text_len, img_len], dim=1) 165 | img_rows_text_cols, img_rows_img_cols = img_rows.split([text_len, img_len], dim=1) 166 | 167 | double_regional_mask = torch.cat([ 168 | torch.cat([img_rows_img_cols, img_rows_text_cols], dim=1), 169 | torch.cat([text_rows_img_cols, text_rows_text_cols], dim=1) 170 | ], dim=0) 171 | 172 | # Patch 173 | double_regional_mask = RegionalMask(double_regional_mask, start_percent, end_percent) 174 | single_regional_mask = RegionalMask(regional_mask, start_percent, end_percent) 175 | regional_conditioning = RegionalConditioning(regional_conditioning, start_percent, end_percent, main_cond_strength, always_included) 176 | 177 | 178 | model.set_model_patch(regional_conditioning, 'regional_conditioning') 179 | 180 | for block_idx in attn_override['double']: 181 | model.set_model_patch_replace(double_regional_mask, f"double", "mask_fn", int(block_idx)) 182 | 183 | for block_idx in attn_override['single']: 184 | model.set_model_patch_replace(single_regional_mask, f"single", "mask_fn", int(block_idx)) 185 | 186 | return (model,) 187 | -------------------------------------------------------------------------------- /nodes/flowedit_nodes.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from tqdm import trange 3 | 4 | from comfy.samplers import KSAMPLER, CFGGuider, sampling_function 5 | 6 | 7 | class FlowEditGuider(CFGGuider): 8 | def __init__(self, model_patcher): 9 | super().__init__(model_patcher) 10 | self.cfgs = {} 11 | self.num_repeats = 1 12 | 13 | def set_conds(self, **kwargs): 14 | self.inner_set_conds(kwargs) 15 | 16 | def set_cfgs(self, **kwargs): 17 | self.cfgs = {**kwargs} 18 | 19 | def set_num_repeats(self, num_repeats): 20 | self.num_repeats = num_repeats 21 | 22 | def predict_noise(self, x, timestep, model_options={}, seed=None): 23 | latent_type = model_options['transformer_options']['latent_type'] 24 | positive = self.conds.get(f'{latent_type}_positive', None) 25 | negative = self.conds.get(f'{latent_type}_negative', None) 26 | cfg = self.cfgs.get(latent_type, self.cfg) 27 | 28 | if self.num_repeats == 1: 29 | return sampling_function(self.inner_model, x, timestep, negative, positive, cfg, model_options=model_options, seed=seed) 30 | 31 | # Multiple samples case 32 | predictions = None 33 | for i in range(self.num_repeats): 34 | current_seed = None if seed is None else seed + i 35 | current_pred = sampling_function( 36 | self.inner_model, x, timestep, negative, positive, cfg, 37 | model_options=model_options, seed=current_seed 38 | ) 39 | if predictions is None: 40 | predictions = current_pred 41 | else: 42 | predictions += current_pred 43 | 44 | return predictions / self.num_repeats 45 | 46 | 47 | # Basic node for model compatibility 48 | class HYFlowEditGuiderNode: 49 | @classmethod 50 | def INPUT_TYPES(s): 51 | return {"required": 52 | { 53 | "model": ("MODEL",), 54 | "source_cond": ("CONDITIONING", ), 55 | "target_cond": ("CONDITIONING", ), 56 | } 57 | } 58 | 59 | RETURN_TYPES = ("GUIDER",) 60 | FUNCTION = "get_guider" 61 | CATEGORY = "hunyuanloom" 62 | 63 | def get_guider(self, model, source_cond, target_cond): 64 | guider = FlowEditGuider(model) 65 | guider.set_conds(source_positive=source_cond, target_positive=target_cond) 66 | guider.set_cfg(1.0) 67 | return (guider,) 68 | 69 | class HYFlowEditGuiderAdvNode: 70 | @classmethod 71 | def INPUT_TYPES(s): 72 | return {"required": 73 | { 74 | "model": ("MODEL",), 75 | "source_cond": ("CONDITIONING", ), 76 | "target_cond": ("CONDITIONING", ), 77 | "num_repeats": ("INT", {"default": 1, "min": 1, "max": 10}), 78 | } 79 | } 80 | 81 | RETURN_TYPES = ("GUIDER",) 82 | FUNCTION = "get_guider" 83 | CATEGORY = "hunyuanloom" 84 | 85 | def get_guider(self, model, source_cond, target_cond, num_repeats): 86 | guider = FlowEditGuider(model) 87 | guider.set_conds(source_positive=source_cond, target_positive=target_cond) 88 | guider.set_cfg(1.0) 89 | guider.set_num_repeats(num_repeats) 90 | return (guider,) 91 | 92 | 93 | class HYFlowEditGuiderCFGNode: 94 | @classmethod 95 | def INPUT_TYPES(s): 96 | return {"required": 97 | { 98 | "model": ("MODEL",), 99 | "source_cond": ("CONDITIONING", ), 100 | "source_uncond": ("CONDITIONING", ), 101 | "target_cond": ("CONDITIONING", ), 102 | "target_uncond": ("CONDITIONING", ), 103 | "source_cfg": ("FLOAT", {"default": 2, "min": 0, "max": 0xffffffffffffffff, "step": 0.01 }), 104 | "target_cfg": ("FLOAT", {"default": 4.5, "min": 0, "max": 0xffffffffffffffff, "step": 0.01 }), 105 | } 106 | } 107 | 108 | RETURN_TYPES = ("GUIDER",) 109 | FUNCTION = "get_guider" 110 | CATEGORY = "hunyuanloom" 111 | 112 | def get_guider(self, model, source_cond, source_uncond, target_cond, target_uncond, source_cfg, target_cfg): 113 | guider = FlowEditGuider(model) 114 | guider.set_conds(source_positive=source_cond, source_negative=source_uncond, 115 | target_positive=target_cond, target_negative=target_uncond) 116 | guider.set_cfgs(source=source_cfg, target=target_cfg) 117 | return (guider,) 118 | 119 | 120 | class HYFlowEditGuiderCFGAdvNode: 121 | @classmethod 122 | def INPUT_TYPES(s): 123 | return {"required": 124 | { 125 | "model": ("MODEL",), 126 | "source_cond": ("CONDITIONING", ), 127 | "source_uncond": ("CONDITIONING", ), 128 | "target_cond": ("CONDITIONING", ), 129 | "target_uncond": ("CONDITIONING", ), 130 | "source_cfg": ("FLOAT", {"default": 2, "min": 0, "max": 0xffffffffffffffff, "step": 0.01 }), 131 | "target_cfg": ("FLOAT", {"default": 4.5, "min": 0, "max": 0xffffffffffffffff, "step": 0.01 }), 132 | "num_repeats": ("INT", {"default": 1, "min": 1, "max": 10}), 133 | } 134 | } 135 | 136 | RETURN_TYPES = ("GUIDER",) 137 | FUNCTION = "get_guider" 138 | CATEGORY = "hunyuanloom" 139 | 140 | def get_guider(self, model, source_cond, source_uncond, target_cond, target_uncond, source_cfg, target_cfg, num_repeats): 141 | guider = FlowEditGuider(model) 142 | guider.set_conds(source_positive=source_cond, source_negative=source_uncond, 143 | target_positive=target_cond, target_negative=target_uncond) 144 | guider.set_cfgs(source=source_cfg, target=target_cfg) 145 | guider.set_num_repeats(num_repeats) 146 | return (guider,) 147 | 148 | def get_flowedit_sample(skip_steps, refine_steps, generator): 149 | @torch.no_grad() 150 | def flowedit_sample(model, x_init, sigmas, extra_args=None, callback=None, disable=None): 151 | extra_args = {} if extra_args is None else extra_args 152 | 153 | model_options = extra_args.get('model_options', {}) 154 | transformer_options = model_options.get('transformer_options', {}) 155 | transformer_options = {**transformer_options} 156 | model_options['transformer_options'] = transformer_options 157 | extra_args['model_options'] = model_options 158 | 159 | source_extra_args = {**extra_args, 'model_options': { 'transformer_options': { 'latent_type': 'source'} }} 160 | 161 | sigmas = sigmas[skip_steps:] 162 | 163 | x_tgt = x_init.clone() 164 | N = len(sigmas)-1 165 | s_in = x_init.new_ones([x_init.shape[0]]) 166 | noise_mask = extra_args.get('denoise_mask', None) 167 | if noise_mask is None: 168 | noise_mask = torch.ones_like(x_init) 169 | else: 170 | extra_args['denoise_mask'] = None 171 | source_extra_args['denoise_mask'] = None 172 | 173 | for i in trange(N, disable=disable): 174 | sigma = sigmas[i] 175 | noise = torch.randn(x_init.shape, generator=generator).to(x_init.device) 176 | 177 | zt_src = (1-sigma)*x_init + sigma*noise 178 | 179 | if i < N-refine_steps: 180 | zt_tgt = x_tgt + zt_src - x_init 181 | vt_src = model(zt_src, sigma*s_in, **source_extra_args) 182 | else: 183 | if i == N-refine_steps: 184 | zt_tgt = x_tgt + (zt_src - x_init) 185 | x_tgt = x_tgt + (zt_src - x_init) * noise_mask 186 | else: 187 | zt_tgt = x_tgt * (noise_mask) + (1-noise_mask) * ( (1-sigma)*x_tgt + sigma*noise ) 188 | vt_src = 0 189 | 190 | transformer_options['latent_type'] = 'target' 191 | vt_tgt = model(zt_tgt, sigma*s_in, **extra_args) 192 | 193 | v_delta = vt_tgt - vt_src 194 | x_tgt += (sigmas[i+1] - sigmas[i]) * v_delta * noise_mask 195 | 196 | if callback is not None: 197 | callback({'x': x_tgt, 'denoised': x_tgt, 'i': i+skip_steps, 'sigma': sigmas[i], 'sigma_hat': sigmas[i]}) 198 | 199 | return x_tgt 200 | 201 | return flowedit_sample 202 | 203 | 204 | class HYFlowEditSamplerNode: 205 | @classmethod 206 | def INPUT_TYPES(s): 207 | return {"required": { 208 | "skip_steps": ("INT", {"default": 4, "min": 0, "max": 0xffffffffffffffff }), 209 | "drift_steps": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff }), 210 | "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff }), 211 | }, "optional": { 212 | }} 213 | RETURN_TYPES = ("SAMPLER",) 214 | FUNCTION = "build" 215 | 216 | CATEGORY = "hunyuanloom" 217 | 218 | def build(self, skip_steps, drift_steps, seed): 219 | generator = torch.manual_seed(seed) 220 | sampler = KSAMPLER(get_flowedit_sample(skip_steps, drift_steps, generator)) 221 | return (sampler, ) 222 | -------------------------------------------------------------------------------- /nodes/flow_edit_nodes.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import gc 3 | 4 | 5 | from diffusers.utils.torch_utils import randn_tensor 6 | import comfy.model_management as mm 7 | from ..utils.rope_utils import get_rotary_pos_embed 8 | from ..utils.latent_preview import prepare_callback 9 | 10 | 11 | class HyVideoFlowEditSamplerNode: 12 | @classmethod 13 | def INPUT_TYPES(s): 14 | return { 15 | "required": { 16 | "model": ("HYVIDEOMODEL",), 17 | "source_embeds": ("HYVIDEMBEDS", ), 18 | "target_embeds": ("HYVIDEMBEDS", ), 19 | "samples": ("LATENT", {"tooltip": "init Latents to use for video2video process"} ), 20 | "steps": ("INT", {"default": 30, "min": 1}), 21 | "skip_steps": ("INT", {"default": 4, "min": 0}), 22 | "drift_steps": ("INT", {"default": 0, "min": 0}), 23 | "source_guidance_scale": ("FLOAT", {"default": 6.0, "min": 0.0, "max": 30.0, "step": 0.01}), 24 | "target_guidance_scale": ("FLOAT", {"default": 12.0, "min": 0.0, "max": 30.0, "step": 0.01}), 25 | "flow_shift": ("FLOAT", {"default": 3.0, "min": 1.0, "max": 30.0, "step": 0.01}), 26 | "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}), 27 | "force_offload": ("BOOLEAN", {"default": True}), 28 | }, 29 | } 30 | 31 | RETURN_TYPES = ("LATENT",) 32 | RETURN_NAMES = ("samples",) 33 | FUNCTION = "process" 34 | CATEGORY = "hunyuanloom" 35 | 36 | def process(self, 37 | model, 38 | source_embeds, 39 | target_embeds, 40 | flow_shift, 41 | steps, 42 | skip_steps, 43 | drift_steps, 44 | source_guidance_scale, 45 | target_guidance_scale, 46 | seed, 47 | samples, 48 | force_offload): 49 | model = model.model 50 | device = mm.get_torch_device() 51 | offload_device = mm.unet_offload_device() 52 | dtype = model["dtype"] 53 | transformer = model["pipe"].transformer 54 | pipeline = model["pipe"] 55 | 56 | generator = torch.Generator(device=torch.device("cpu")).manual_seed(seed) 57 | 58 | latents = samples["samples"] if samples is not None else None 59 | batch_size, num_channels_latents, latent_num_frames, latent_height, latent_width = latents.shape 60 | height = latent_height * pipeline.vae_scale_factor 61 | width = latent_width * pipeline.vae_scale_factor 62 | num_frames = (latent_num_frames - 1) * 4 + 1 63 | 64 | if width <= 0 or height <= 0 or num_frames <= 0: 65 | raise ValueError( 66 | f"`height` and `width` and `video_length` must be positive integers, got height={height}, width={width}, video_length={num_frames}" 67 | ) 68 | if (num_frames - 1) % 4 != 0: 69 | raise ValueError( 70 | f"`video_length-1` must be a multiple of 4, got {num_frames}" 71 | ) 72 | 73 | freqs_cos, freqs_sin = get_rotary_pos_embed(transformer, num_frames, height, width) 74 | 75 | pipeline.scheduler.shift = flow_shift 76 | 77 | if model["block_swap_args"] is not None: 78 | for name, param in transformer.named_parameters(): 79 | #print(name, param.data.device) 80 | if "single" not in name and "double" not in name: 81 | param.data = param.data.to(device) 82 | 83 | transformer.block_swap( 84 | model["block_swap_args"]["double_blocks_to_swap"] - 1 , 85 | model["block_swap_args"]["single_blocks_to_swap"] - 1, 86 | offload_txt_in = model["block_swap_args"]["offload_txt_in"], 87 | offload_img_in = model["block_swap_args"]["offload_img_in"], 88 | ) 89 | elif model["manual_offloading"]: 90 | transformer.to(device) 91 | 92 | mm.soft_empty_cache() 93 | gc.collect() 94 | 95 | try: 96 | torch.cuda.reset_peak_memory_stats(device) 97 | except: 98 | pass 99 | 100 | pipeline.scheduler.set_timesteps(steps, device=device) 101 | timesteps = pipeline.scheduler.timesteps 102 | timesteps = torch.cat([timesteps, torch.tensor([0]).to(timesteps.device)]).to(timesteps.device) 103 | 104 | latent_video_length = (num_frames - 1) // 4 + 1 105 | 106 | # 5. Prepare latent variables 107 | num_channels_latents = transformer.config.in_channels 108 | 109 | latents = latents.to(device) 110 | 111 | shape = ( 112 | 1, 113 | num_channels_latents, 114 | latent_video_length, 115 | int(height) // pipeline.vae_scale_factor, 116 | int(width) // pipeline.vae_scale_factor, 117 | ) 118 | noise = randn_tensor(shape, generator=generator, device=device, dtype=torch.float32) 119 | 120 | frames_needed = noise.shape[1] 121 | current_frames = latents.shape[1] 122 | 123 | if frames_needed > current_frames: 124 | repeat_factor = frames_needed - current_frames 125 | additional_frame = torch.randn((latents.size(0), repeat_factor, latents.size(2), latents.size(3), latents.size(4)), dtype=latents.dtype, device=latents.device) 126 | latents = torch.cat((additional_frame, latents), dim=1) 127 | self.additional_frames = repeat_factor 128 | elif frames_needed < current_frames: 129 | latents = latents[:, :frames_needed, :, :, :] 130 | 131 | 132 | # 7. Denoising loop 133 | self._num_timesteps = len(timesteps) 134 | 135 | callback = prepare_callback(transformer, steps) 136 | 137 | x_init = latents.clone() 138 | 139 | from comfy.utils import ProgressBar 140 | from tqdm import tqdm 141 | comfy_pbar = ProgressBar(len(timesteps)) 142 | N = len(timesteps) 143 | 144 | x_tgt = latents 145 | 146 | with tqdm(total=len(timesteps)) as progress_bar: 147 | for idx, (t, t_prev) in enumerate(zip(timesteps[:-1], timesteps[1:])): 148 | if idx < skip_steps: 149 | continue 150 | t_expand = t.repeat(x_init.shape[0]) 151 | source_guidance_expand = ( 152 | torch.tensor( 153 | [source_guidance_scale] * x_init.shape[0], 154 | dtype=torch.float32, 155 | device=device, 156 | ).to(pipeline.base_dtype) 157 | * 1000.0 158 | if source_guidance_scale is not None 159 | else None 160 | ) 161 | 162 | target_guidance_expand = ( 163 | torch.tensor( 164 | [target_guidance_scale] * x_init.shape[0], 165 | dtype=torch.float32, 166 | device=device, 167 | ).to(pipeline.base_dtype) 168 | * 1000.0 169 | if target_guidance_scale is not None 170 | else None 171 | ) 172 | 173 | # predict the noise residual 174 | with torch.autocast( 175 | device_type="cuda", dtype=pipeline.base_dtype, enabled=True 176 | ): 177 | 178 | noise = torch.randn(x_init.shape, generator=generator).to(x_init.device) 179 | 180 | sigma = t / 1000.0 181 | sigma_prev = t_prev / 1000.0 182 | 183 | zt_src = (1-sigma)*x_init + sigma*noise 184 | 185 | if idx < N-drift_steps: 186 | zt_tgt = x_tgt + zt_src - x_init 187 | vt_src = transformer( # For an input image (129, 192, 336) (1, 256, 256) 188 | zt_src, # [2, 16, 33, 24, 42] 189 | t_expand, # [2] 190 | text_states=source_embeds["prompt_embeds"], # [2, 256, 4096] 191 | text_mask=source_embeds["attention_mask"], # [2, 256] 192 | text_states_2=source_embeds["prompt_embeds_2"], # [2, 768] 193 | freqs_cos=freqs_cos, # [seqlen, head_dim] 194 | freqs_sin=freqs_sin, # [seqlen, head_dim] 195 | guidance=source_guidance_expand, 196 | stg_block_idx=-1, 197 | stg_mode=None, 198 | return_dict=True, 199 | )["x"] 200 | else: 201 | if idx == N - drift_steps: 202 | x_tgt = x_tgt + zt_src - x_init 203 | zt_tgt = x_tgt 204 | vt_src = 0 205 | 206 | vt_tgt = transformer( # For an input image (129, 192, 336) (1, 256, 256) 207 | zt_tgt, # [2, 16, 33, 24, 42] 208 | t_expand, # [2] 209 | text_states=target_embeds["prompt_embeds"], # [2, 256, 4096] 210 | text_mask=target_embeds["attention_mask"], # [2, 256] 211 | text_states_2=target_embeds["prompt_embeds_2"], # [2, 768] 212 | freqs_cos=freqs_cos, # [seqlen, head_dim] 213 | freqs_sin=freqs_sin, # [seqlen, head_dim] 214 | guidance=target_guidance_expand, 215 | stg_block_idx=-1, 216 | stg_mode=None, 217 | return_dict=True, 218 | )["x"] 219 | 220 | v_delta = vt_tgt - vt_src 221 | 222 | x_tgt = x_tgt.to(torch.float32) 223 | v_delta = v_delta.to(torch.float32) 224 | 225 | 226 | x_tgt = x_tgt + (sigma_prev - sigma) * v_delta 227 | x_tgt = x_tgt.to(torch.bfloat16) 228 | 229 | # compute the previous noisy sample x_t -> x_t-1 230 | #latents = pipeline.scheduler.step(noise_pred, t, latents, return_dict=False)[0] 231 | 232 | 233 | progress_bar.update() 234 | if callback is not None: 235 | callback(idx, latents.detach()[-1].permute(1,0,2,3), None, steps) 236 | else: 237 | comfy_pbar.update(1) 238 | 239 | try: 240 | torch.cuda.reset_peak_memory_stats(device) 241 | except: 242 | pass 243 | 244 | if force_offload: 245 | if model["manual_offloading"]: 246 | transformer.to(offload_device) 247 | mm.soft_empty_cache() 248 | gc.collect() 249 | 250 | return ({ 251 | "samples": x_tgt 252 | },) 253 | 254 | 255 | -------------------------------------------------------------------------------- /nodes/wrapper_flow_edit_nodes.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import gc 3 | 4 | 5 | from diffusers.utils.torch_utils import randn_tensor 6 | import comfy.model_management as mm 7 | from ..utils.rope_utils import get_rotary_pos_embed 8 | from ..utils.latent_preview import prepare_callback 9 | 10 | VAE_SCALING_FACTOR = 0.476986 11 | 12 | class HyVideoFlowEditSamplerNode: 13 | @classmethod 14 | def INPUT_TYPES(s): 15 | return { 16 | "required": { 17 | "model": ("HYVIDEOMODEL",), 18 | "source_embeds": ("HYVIDEMBEDS", ), 19 | "target_embeds": ("HYVIDEMBEDS", ), 20 | "samples": ("LATENT", {"tooltip": "init Latents to use for video2video process"} ), 21 | "steps": ("INT", {"default": 30, "min": 1}), 22 | "skip_steps": ("INT", {"default": 4, "min": 0}), 23 | "drift_steps": ("INT", {"default": 0, "min": 0}), 24 | "source_guidance_scale": ("FLOAT", {"default": 6.0, "min": 0.0, "max": 30.0, "step": 0.01}), 25 | "target_guidance_scale": ("FLOAT", {"default": 12.0, "min": 0.0, "max": 30.0, "step": 0.01}), 26 | "drift_guidance_scale": ("FLOAT", {"default": 6.0, "min": 0.0, "max": 30.0, "step": 0.01}), 27 | "flow_shift": ("FLOAT", {"default": 6.0, "min": 1.0, "max": 30.0, "step": 0.01}), 28 | "drift_flow_shift": ("FLOAT", {"default": 3.0, "min": 1.0, "max": 30.0, "step": 0.01}), 29 | "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}), 30 | "force_offload": ("BOOLEAN", {"default": True}), 31 | }, 32 | } 33 | 34 | RETURN_TYPES = ("LATENT",) 35 | RETURN_NAMES = ("samples",) 36 | FUNCTION = "process" 37 | CATEGORY = "hunyuanloom" 38 | 39 | def process(self, 40 | model, 41 | source_embeds, 42 | target_embeds, 43 | flow_shift, 44 | drift_flow_shift, 45 | steps, 46 | skip_steps, 47 | drift_steps, 48 | source_guidance_scale, 49 | target_guidance_scale, 50 | drift_guidance_scale, 51 | seed, 52 | samples, 53 | force_offload): 54 | model = model.model 55 | 56 | device = mm.get_torch_device() 57 | offload_device = mm.unet_offload_device() 58 | transformer = model["pipe"].transformer 59 | pipeline = model["pipe"] 60 | 61 | generator = torch.Generator(device=torch.device("cpu")).manual_seed(seed) 62 | 63 | latents = samples["samples"] * VAE_SCALING_FACTOR if samples is not None else None 64 | batch_size, num_channels_latents, latent_num_frames, latent_height, latent_width = latents.shape 65 | height = latent_height * pipeline.vae_scale_factor 66 | width = latent_width * pipeline.vae_scale_factor 67 | num_frames = (latent_num_frames - 1) * 4 + 1 68 | 69 | if width <= 0 or height <= 0 or num_frames <= 0: 70 | raise ValueError( 71 | f"`height` and `width` and `video_length` must be positive integers, got height={height}, width={width}, video_length={num_frames}" 72 | ) 73 | if (num_frames - 1) % 4 != 0: 74 | raise ValueError( 75 | f"`video_length-1` must be a multiple of 4, got {num_frames}" 76 | ) 77 | 78 | freqs_cos, freqs_sin = get_rotary_pos_embed(transformer, num_frames, height, width) 79 | freqs_cos = freqs_cos.to(device) 80 | freqs_sin = freqs_sin.to(device) 81 | 82 | if model["block_swap_args"] is not None: 83 | for name, param in transformer.named_parameters(): 84 | #print(name, param.data.device) 85 | if "single" not in name and "double" not in name: 86 | param.data = param.data.to(device) 87 | 88 | transformer.block_swap( 89 | model["block_swap_args"]["double_blocks_to_swap"] - 1 , 90 | model["block_swap_args"]["single_blocks_to_swap"] - 1, 91 | offload_txt_in = model["block_swap_args"]["offload_txt_in"], 92 | offload_img_in = model["block_swap_args"]["offload_img_in"], 93 | ) 94 | elif model["manual_offloading"]: 95 | transformer.to(device) 96 | 97 | mm.soft_empty_cache() 98 | gc.collect() 99 | 100 | try: 101 | torch.cuda.reset_peak_memory_stats(device) 102 | except: 103 | pass 104 | 105 | # drift_flow_shift 106 | 107 | pipeline.scheduler.flow_shift = flow_shift 108 | pipeline.scheduler.set_timesteps(steps, device=device) 109 | timesteps = pipeline.scheduler.timesteps 110 | timesteps = torch.cat([timesteps, torch.tensor([0]).to(timesteps.device)]).to(timesteps.device) 111 | 112 | pipeline.scheduler.flow_shift = drift_flow_shift 113 | pipeline.scheduler.set_timesteps(steps, device=device) 114 | drift_timesteps = pipeline.scheduler.timesteps 115 | drift_timesteps = torch.cat([drift_timesteps, torch.tensor([0]).to(drift_timesteps.device)]).to(drift_timesteps.device) 116 | 117 | timesteps[-drift_steps:] = drift_timesteps[-drift_steps:] 118 | 119 | latent_video_length = (num_frames - 1) // 4 + 1 120 | 121 | # 5. Prepare latent variables 122 | num_channels_latents = transformer.config.in_channels 123 | 124 | latents = latents.to(device) 125 | 126 | shape = ( 127 | 1, 128 | num_channels_latents, 129 | latent_video_length, 130 | int(height) // pipeline.vae_scale_factor, 131 | int(width) // pipeline.vae_scale_factor, 132 | ) 133 | noise = randn_tensor(shape, generator=generator, device=device, dtype=torch.float32) 134 | 135 | frames_needed = noise.shape[1] 136 | current_frames = latents.shape[1] 137 | 138 | if frames_needed > current_frames: 139 | repeat_factor = frames_needed - current_frames 140 | additional_frame = torch.randn((latents.size(0), repeat_factor, latents.size(2), latents.size(3), latents.size(4)), dtype=latents.dtype, device=latents.device) 141 | latents = torch.cat((additional_frame, latents), dim=1) 142 | self.additional_frames = repeat_factor 143 | elif frames_needed < current_frames: 144 | latents = latents[:, :frames_needed, :, :, :] 145 | 146 | 147 | # 7. Denoising loop 148 | callback = prepare_callback(transformer, steps) 149 | 150 | x_init = latents.clone() 151 | 152 | from comfy.utils import ProgressBar 153 | from tqdm import tqdm 154 | comfy_pbar = ProgressBar(len(timesteps)) 155 | N = len(timesteps) 156 | 157 | x_tgt = latents 158 | 159 | with tqdm(total=len(timesteps)) as progress_bar: 160 | for idx, (t, t_prev) in enumerate(zip(timesteps[:-1], timesteps[1:])): 161 | if idx < skip_steps: 162 | continue 163 | t_expand = t.repeat(x_init.shape[0]) 164 | source_guidance_expand = ( 165 | torch.tensor( 166 | [source_guidance_scale] * x_init.shape[0], 167 | dtype=torch.float32, 168 | device=device, 169 | ).to(pipeline.base_dtype) 170 | * 1000.0 171 | if source_guidance_scale is not None 172 | else None 173 | ) 174 | 175 | if idx < N-drift_steps: 176 | current_guidance_scale = target_guidance_scale 177 | else: 178 | current_guidance_scale = drift_guidance_scale 179 | 180 | target_guidance_expand = ( 181 | torch.tensor( 182 | [current_guidance_scale] * x_init.shape[0], 183 | dtype=torch.float32, 184 | device=device, 185 | ).to(pipeline.base_dtype) 186 | * 1000.0 187 | if target_guidance_scale is not None 188 | else None 189 | ) 190 | 191 | with torch.autocast( 192 | device_type="cuda", dtype=pipeline.base_dtype, enabled=True 193 | ): 194 | noise = torch.randn(x_init.shape, generator=generator).to(x_init.device) 195 | 196 | sigma = t / 1000.0 197 | sigma_prev = t_prev / 1000.0 198 | 199 | zt_src = (1-sigma) * x_init + sigma * noise 200 | zt_tgt = x_tgt + zt_src - x_init 201 | 202 | if idx < N-drift_steps: 203 | vt_src = transformer( # For an input image (129, 192, 336) (1, 256, 256) 204 | zt_src, # [2, 16, 33, 24, 42] 205 | t_expand, # [2] 206 | text_states=source_embeds["prompt_embeds"], # [2, 256, 4096] 207 | text_mask=source_embeds["attention_mask"], # [2, 256] 208 | text_states_2=source_embeds["prompt_embeds_2"], # [2, 768] 209 | freqs_cos=freqs_cos, # [seqlen, head_dim] 210 | freqs_sin=freqs_sin, # [seqlen, head_dim] 211 | guidance=source_guidance_expand, 212 | stg_block_idx=-1, 213 | stg_mode=None, 214 | return_dict=True, 215 | )["x"] 216 | else: 217 | if idx == N - drift_steps: 218 | x_tgt = zt_tgt 219 | zt_tgt = x_tgt 220 | vt_src = 0 221 | 222 | vt_tgt = transformer( # For an input image (129, 192, 336) (1, 256, 256) 223 | zt_tgt, # [2, 16, 33, 24, 42] 224 | t_expand, # [2] 225 | text_states=target_embeds["prompt_embeds"], # [2, 256, 4096] 226 | text_mask=target_embeds["attention_mask"], # [2, 256] 227 | text_states_2=target_embeds["prompt_embeds_2"], # [2, 768] 228 | freqs_cos=freqs_cos, # [seqlen, head_dim] 229 | freqs_sin=freqs_sin, # [seqlen, head_dim] 230 | guidance=target_guidance_expand, 231 | stg_block_idx=-1, 232 | stg_mode=None, 233 | return_dict=True, 234 | )["x"] 235 | 236 | v_delta = vt_tgt - vt_src 237 | 238 | x_tgt = x_tgt.to(torch.float32) 239 | v_delta = v_delta.to(torch.float32) 240 | 241 | x_tgt = x_tgt + (sigma_prev - sigma) * v_delta 242 | x_tgt = x_tgt.to(torch.bfloat16) 243 | 244 | progress_bar.update() 245 | if callback is not None: 246 | callback(idx, (zt_tgt - vt_tgt * sigma).detach()[-1].permute(1,0,2,3), None, steps) 247 | else: 248 | comfy_pbar.update(1) 249 | 250 | try: 251 | torch.cuda.reset_peak_memory_stats(device) 252 | except: 253 | pass 254 | 255 | if force_offload: 256 | if model["manual_offloading"]: 257 | transformer.to(offload_device) 258 | mm.soft_empty_cache() 259 | gc.collect() 260 | 261 | return ({ 262 | "samples": x_tgt / VAE_SCALING_FACTOR 263 | },) 264 | 265 | 266 | -------------------------------------------------------------------------------- /utils/rope_utils.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | import torch 4 | from typing import Union, Tuple, List 5 | 6 | 7 | def _to_tuple(x, dim=2): 8 | if isinstance(x, int): 9 | return (x,) * dim 10 | elif len(x) == dim: 11 | return x 12 | else: 13 | raise ValueError(f"Expected length {dim} or int, but got {x}") 14 | 15 | 16 | def get_meshgrid_nd(start, *args, dim=2): 17 | """ 18 | Get n-D meshgrid with start, stop and num. 19 | 20 | Args: 21 | start (int or tuple): If len(args) == 0, start is num; If len(args) == 1, start is start, args[0] is stop, 22 | step is 1; If len(args) == 2, start is start, args[0] is stop, args[1] is num. For n-dim, start/stop/num 23 | should be int or n-tuple. If n-tuple is provided, the meshgrid will be stacked following the dim order in 24 | n-tuples. 25 | *args: See above. 26 | dim (int): Dimension of the meshgrid. Defaults to 2. 27 | 28 | Returns: 29 | grid (np.ndarray): [dim, ...] 30 | """ 31 | if len(args) == 0: 32 | # start is grid_size 33 | num = _to_tuple(start, dim=dim) 34 | start = (0,) * dim 35 | stop = num 36 | elif len(args) == 1: 37 | # start is start, args[0] is stop, step is 1 38 | start = _to_tuple(start, dim=dim) 39 | stop = _to_tuple(args[0], dim=dim) 40 | num = [stop[i] - start[i] for i in range(dim)] 41 | elif len(args) == 2: 42 | # start is start, args[0] is stop, args[1] is num 43 | start = _to_tuple(start, dim=dim) # Left-Top eg: 12,0 44 | stop = _to_tuple(args[0], dim=dim) # Right-Bottom eg: 20,32 45 | num = _to_tuple(args[1], dim=dim) # Target Size eg: 32,124 46 | else: 47 | raise ValueError(f"len(args) should be 0, 1 or 2, but got {len(args)}") 48 | 49 | # PyTorch implement of np.linspace(start[i], stop[i], num[i], endpoint=False) 50 | axis_grid = [] 51 | for i in range(dim): 52 | a, b, n = start[i], stop[i], num[i] 53 | g = torch.linspace(a, b, n + 1, dtype=torch.float32)[:n] 54 | axis_grid.append(g) 55 | grid = torch.meshgrid(*axis_grid, indexing="ij") # dim x [W, H, D] 56 | grid = torch.stack(grid, dim=0) # [dim, W, H, D] 57 | 58 | return grid 59 | 60 | 61 | ################################################################################# 62 | # Rotary Positional Embedding Functions # 63 | ################################################################################# 64 | # https://github.com/meta-llama/llama/blob/be327c427cc5e89cc1d3ab3d3fec4484df771245/llama/model.py#L80 65 | 66 | 67 | def reshape_for_broadcast( 68 | freqs_cis: Union[torch.Tensor, Tuple[torch.Tensor]], 69 | x: torch.Tensor, 70 | head_first=False, 71 | ): 72 | """ 73 | Reshape frequency tensor for broadcasting it with another tensor. 74 | 75 | This function reshapes the frequency tensor to have the same shape as the target tensor 'x' 76 | for the purpose of broadcasting the frequency tensor during element-wise operations. 77 | 78 | Notes: 79 | When using FlashMHAModified, head_first should be False. 80 | When using Attention, head_first should be True. 81 | 82 | Args: 83 | freqs_cis (Union[torch.Tensor, Tuple[torch.Tensor]]): Frequency tensor to be reshaped. 84 | x (torch.Tensor): Target tensor for broadcasting compatibility. 85 | head_first (bool): head dimension first (except batch dim) or not. 86 | 87 | Returns: 88 | torch.Tensor: Reshaped frequency tensor. 89 | 90 | Raises: 91 | AssertionError: If the frequency tensor doesn't match the expected shape. 92 | AssertionError: If the target tensor 'x' doesn't have the expected number of dimensions. 93 | """ 94 | ndim = x.ndim 95 | assert 0 <= 1 < ndim 96 | 97 | if isinstance(freqs_cis, tuple): 98 | # freqs_cis: (cos, sin) in real space 99 | if head_first: 100 | assert freqs_cis[0].shape == ( 101 | x.shape[-2], 102 | x.shape[-1], 103 | ), f"freqs_cis shape {freqs_cis[0].shape} does not match x shape {x.shape}" 104 | shape = [ 105 | d if i == ndim - 2 or i == ndim - 1 else 1 106 | for i, d in enumerate(x.shape) 107 | ] 108 | else: 109 | assert freqs_cis[0].shape == ( 110 | x.shape[1], 111 | x.shape[-1], 112 | ), f"freqs_cis shape {freqs_cis[0].shape} does not match x shape {x.shape}" 113 | shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)] 114 | return freqs_cis[0].view(*shape), freqs_cis[1].view(*shape) 115 | else: 116 | # freqs_cis: values in complex space 117 | if head_first: 118 | assert freqs_cis.shape == ( 119 | x.shape[-2], 120 | x.shape[-1], 121 | ), f"freqs_cis shape {freqs_cis.shape} does not match x shape {x.shape}" 122 | shape = [ 123 | d if i == ndim - 2 or i == ndim - 1 else 1 124 | for i, d in enumerate(x.shape) 125 | ] 126 | else: 127 | assert freqs_cis.shape == ( 128 | x.shape[1], 129 | x.shape[-1], 130 | ), f"freqs_cis shape {freqs_cis.shape} does not match x shape {x.shape}" 131 | shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)] 132 | return freqs_cis.view(*shape) 133 | 134 | 135 | def rotate_half(x): 136 | x_real, x_imag = ( 137 | x.float().reshape(*x.shape[:-1], -1, 2).unbind(-1) 138 | ) # [B, S, H, D//2] 139 | return torch.stack([-x_imag, x_real], dim=-1).flatten(3) 140 | 141 | 142 | def apply_rotary_emb( 143 | xq: torch.Tensor, 144 | xk: torch.Tensor, 145 | freqs_cis: Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]], 146 | head_first: bool = False, 147 | ) -> Tuple[torch.Tensor, torch.Tensor]: 148 | """ 149 | Apply rotary embeddings to input tensors using the given frequency tensor. 150 | 151 | This function applies rotary embeddings to the given query 'xq' and key 'xk' tensors using the provided 152 | frequency tensor 'freqs_cis'. The input tensors are reshaped as complex numbers, and the frequency tensor 153 | is reshaped for broadcasting compatibility. The resulting tensors contain rotary embeddings and are 154 | returned as real tensors. 155 | 156 | Args: 157 | xq (torch.Tensor): Query tensor to apply rotary embeddings. [B, S, H, D] 158 | xk (torch.Tensor): Key tensor to apply rotary embeddings. [B, S, H, D] 159 | freqs_cis (torch.Tensor or tuple): Precomputed frequency tensor for complex exponential. 160 | head_first (bool): head dimension first (except batch dim) or not. 161 | 162 | Returns: 163 | Tuple[torch.Tensor, torch.Tensor]: Tuple of modified query tensor and key tensor with rotary embeddings. 164 | 165 | """ 166 | xk_out = None 167 | if isinstance(freqs_cis, tuple): 168 | cos, sin = reshape_for_broadcast(freqs_cis, xq, head_first) # [S, D] 169 | cos, sin = cos.to(xq.device), sin.to(xq.device) 170 | # real * cos - imag * sin 171 | # imag * cos + real * sin 172 | xq_out = (xq.float() * cos + rotate_half(xq.float()) * sin).type_as(xq) 173 | xk_out = (xk.float() * cos + rotate_half(xk.float()) * sin).type_as(xk) 174 | else: 175 | # view_as_complex will pack [..., D/2, 2](real) to [..., D/2](complex) 176 | xq_ = torch.view_as_complex( 177 | xq.float().reshape(*xq.shape[:-1], -1, 2) 178 | ) # [B, S, H, D//2] 179 | freqs_cis = reshape_for_broadcast(freqs_cis, xq_, head_first).to( 180 | xq.device 181 | ) # [S, D//2] --> [1, S, 1, D//2] 182 | # (real, imag) * (cos, sin) = (real * cos - imag * sin, imag * cos + real * sin) 183 | # view_as_real will expand [..., D/2](complex) to [..., D/2, 2](real) 184 | xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3).type_as(xq) 185 | xk_ = torch.view_as_complex( 186 | xk.float().reshape(*xk.shape[:-1], -1, 2) 187 | ) # [B, S, H, D//2] 188 | xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3).type_as(xk) 189 | 190 | return xq_out, xk_out 191 | 192 | 193 | def get_nd_rotary_pos_embed( 194 | rope_dim_list, 195 | start, 196 | *args, 197 | theta=10000.0, 198 | use_real=False, 199 | theta_rescale_factor: Union[float, List[float]] = 1.0, 200 | interpolation_factor: Union[float, List[float]] = 1.0, 201 | ): 202 | """ 203 | This is a n-d version of precompute_freqs_cis, which is a RoPE for tokens with n-d structure. 204 | 205 | Args: 206 | rope_dim_list (list of int): Dimension of each rope. len(rope_dim_list) should equal to n. 207 | sum(rope_dim_list) should equal to head_dim of attention layer. 208 | start (int | tuple of int | list of int): If len(args) == 0, start is num; If len(args) == 1, start is start, 209 | args[0] is stop, step is 1; If len(args) == 2, start is start, args[0] is stop, args[1] is num. 210 | *args: See above. 211 | theta (float): Scaling factor for frequency computation. Defaults to 10000.0. 212 | use_real (bool): If True, return real part and imaginary part separately. Otherwise, return complex numbers. 213 | Some libraries such as TensorRT does not support complex64 data type. So it is useful to provide a real 214 | part and an imaginary part separately. 215 | theta_rescale_factor (float): Rescale factor for theta. Defaults to 1.0. 216 | 217 | Returns: 218 | pos_embed (torch.Tensor): [HW, D/2] 219 | """ 220 | 221 | grid = get_meshgrid_nd( 222 | start, *args, dim=len(rope_dim_list) 223 | ) # [3, W, H, D] / [2, W, H] 224 | 225 | if isinstance(theta_rescale_factor, int) or isinstance(theta_rescale_factor, float): 226 | theta_rescale_factor = [theta_rescale_factor] * len(rope_dim_list) 227 | elif isinstance(theta_rescale_factor, list) and len(theta_rescale_factor) == 1: 228 | theta_rescale_factor = [theta_rescale_factor[0]] * len(rope_dim_list) 229 | assert len(theta_rescale_factor) == len( 230 | rope_dim_list 231 | ), "len(theta_rescale_factor) should equal to len(rope_dim_list)" 232 | 233 | if isinstance(interpolation_factor, int) or isinstance(interpolation_factor, float): 234 | interpolation_factor = [interpolation_factor] * len(rope_dim_list) 235 | elif isinstance(interpolation_factor, list) and len(interpolation_factor) == 1: 236 | interpolation_factor = [interpolation_factor[0]] * len(rope_dim_list) 237 | assert len(interpolation_factor) == len( 238 | rope_dim_list 239 | ), "len(interpolation_factor) should equal to len(rope_dim_list)" 240 | 241 | # use 1/ndim of dimensions to encode grid_axis 242 | embs = [] 243 | for i in range(len(rope_dim_list)): 244 | emb = get_1d_rotary_pos_embed( 245 | rope_dim_list[i], 246 | grid[i].reshape(-1), 247 | theta, 248 | use_real=use_real, 249 | theta_rescale_factor=theta_rescale_factor[i], 250 | interpolation_factor=interpolation_factor[i], 251 | ) # 2 x [WHD, rope_dim_list[i]] 252 | embs.append(emb) 253 | 254 | if use_real: 255 | cos = torch.cat([emb[0] for emb in embs], dim=1) # (WHD, D/2) 256 | sin = torch.cat([emb[1] for emb in embs], dim=1) # (WHD, D/2) 257 | return cos, sin 258 | else: 259 | emb = torch.cat(embs, dim=1) # (WHD, D/2) 260 | return emb 261 | 262 | 263 | def get_1d_rotary_pos_embed( 264 | dim: int, 265 | pos: Union[torch.FloatTensor, int], 266 | theta: float = 10000.0, 267 | use_real: bool = False, 268 | theta_rescale_factor: float = 1.0, 269 | interpolation_factor: float = 1.0, 270 | ) -> Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]: 271 | """ 272 | Precompute the frequency tensor for complex exponential (cis) with given dimensions. 273 | (Note: `cis` means `cos + i * sin`, where i is the imaginary unit.) 274 | 275 | This function calculates a frequency tensor with complex exponential using the given dimension 'dim' 276 | and the end index 'end'. The 'theta' parameter scales the frequencies. 277 | The returned tensor contains complex values in complex64 data type. 278 | 279 | Args: 280 | dim (int): Dimension of the frequency tensor. 281 | pos (int or torch.FloatTensor): Position indices for the frequency tensor. [S] or scalar 282 | theta (float, optional): Scaling factor for frequency computation. Defaults to 10000.0. 283 | use_real (bool, optional): If True, return real part and imaginary part separately. 284 | Otherwise, return complex numbers. 285 | theta_rescale_factor (float, optional): Rescale factor for theta. Defaults to 1.0. 286 | 287 | Returns: 288 | freqs_cis: Precomputed frequency tensor with complex exponential. [S, D/2] 289 | freqs_cos, freqs_sin: Precomputed frequency tensor with real and imaginary parts separately. [S, D] 290 | """ 291 | if isinstance(pos, int): 292 | pos = torch.arange(pos).float() 293 | 294 | # proposed by reddit user bloc97, to rescale rotary embeddings to longer sequence length without fine-tuning 295 | # has some connection to NTK literature 296 | if theta_rescale_factor != 1.0: 297 | theta *= theta_rescale_factor ** (dim / (dim - 2)) 298 | 299 | freqs = 1.0 / ( 300 | theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim) 301 | ) # [D/2] 302 | # assert interpolation_factor == 1.0, f"interpolation_factor: {interpolation_factor}" 303 | freqs = torch.outer(pos * interpolation_factor, freqs) # [S, D/2] 304 | if use_real: 305 | freqs_cos = freqs.cos().repeat_interleave(2, dim=1) # [S, D] 306 | freqs_sin = freqs.sin().repeat_interleave(2, dim=1) # [S, D] 307 | return freqs_cos, freqs_sin 308 | else: 309 | freqs_cis = torch.polar( 310 | torch.ones_like(freqs), freqs 311 | ) # complex64 # [S, D/2] 312 | return freqs_cis 313 | 314 | 315 | def get_rotary_pos_embed(transformer, video_length, height, width): 316 | target_ndim = 3 317 | ndim = 5 - 2 318 | rope_theta = 225 319 | patch_size = transformer.patch_size 320 | rope_dim_list = transformer.rope_dim_list 321 | hidden_size = transformer.hidden_size 322 | heads_num = transformer.heads_num 323 | head_dim = hidden_size // heads_num 324 | 325 | # 884 326 | latents_size = [(video_length - 1) // 4 + 1, height // 8, width // 8] 327 | 328 | if isinstance(patch_size, int): 329 | assert all(s % patch_size == 0 for s in latents_size), ( 330 | f"Latent size(last {ndim} dimensions) should be divisible by patch size({patch_size}), " 331 | f"but got {latents_size}." 332 | ) 333 | rope_sizes = [s // patch_size for s in latents_size] 334 | elif isinstance(patch_size, list): 335 | assert all( 336 | s % patch_size[idx] == 0 337 | for idx, s in enumerate(latents_size) 338 | ), ( 339 | f"Latent size(last {ndim} dimensions) should be divisible by patch size({patch_size}), " 340 | f"but got {latents_size}." 341 | ) 342 | rope_sizes = [ 343 | s // patch_size[idx] for idx, s in enumerate(latents_size) 344 | ] 345 | 346 | if len(rope_sizes) != target_ndim: 347 | rope_sizes = [1] * (target_ndim - len(rope_sizes)) + rope_sizes # time axis 348 | 349 | if rope_dim_list is None: 350 | rope_dim_list = [head_dim // target_ndim for _ in range(target_ndim)] 351 | assert ( 352 | sum(rope_dim_list) == head_dim 353 | ), "sum(rope_dim_list) should equal to head_dim of attention layer" 354 | freqs_cos, freqs_sin = get_nd_rotary_pos_embed( 355 | rope_dim_list, 356 | rope_sizes, 357 | theta=rope_theta, 358 | use_real=True, 359 | theta_rescale_factor=1, 360 | ) 361 | return freqs_cos, freqs_sin 362 | -------------------------------------------------------------------------------- /example_workflows/example_hy_flowedit.json: -------------------------------------------------------------------------------- 1 | { 2 | "last_node_id": 89, 3 | "last_link_id": 231, 4 | "nodes": [ 5 | { 6 | "id": 73, 7 | "type": "VAEDecodeTiled", 8 | "pos": [ 9 | 1470.9056396484375, 10 | 171.9037322998047 11 | ], 12 | "size": [ 13 | 210, 14 | 102 15 | ], 16 | "flags": {}, 17 | "order": 16, 18 | "mode": 0, 19 | "inputs": [ 20 | { 21 | "name": "samples", 22 | "type": "LATENT", 23 | "link": 210 24 | }, 25 | { 26 | "name": "vae", 27 | "type": "VAE", 28 | "link": 211 29 | } 30 | ], 31 | "outputs": [ 32 | { 33 | "name": "IMAGE", 34 | "type": "IMAGE", 35 | "links": [ 36 | 216 37 | ], 38 | "slot_index": 0 39 | } 40 | ], 41 | "properties": { 42 | "Node name for S&R": "VAEDecodeTiled" 43 | }, 44 | "widgets_values": [ 45 | 256, 46 | 64 47 | ] 48 | }, 49 | { 50 | "id": 13, 51 | "type": "SamplerCustomAdvanced", 52 | "pos": [ 53 | 1163.367431640625, 54 | 168.47523498535156 55 | ], 56 | "size": [ 57 | 272.3617858886719, 58 | 124.53733825683594 59 | ], 60 | "flags": {}, 61 | "order": 15, 62 | "mode": 0, 63 | "inputs": [ 64 | { 65 | "name": "noise", 66 | "type": "NOISE", 67 | "link": 217, 68 | "slot_index": 0 69 | }, 70 | { 71 | "name": "guider", 72 | "type": "GUIDER", 73 | "link": 218, 74 | "slot_index": 1 75 | }, 76 | { 77 | "name": "sampler", 78 | "type": "SAMPLER", 79 | "link": 231, 80 | "slot_index": 2 81 | }, 82 | { 83 | "name": "sigmas", 84 | "type": "SIGMAS", 85 | "link": 20, 86 | "slot_index": 3 87 | }, 88 | { 89 | "name": "latent_image", 90 | "type": "LATENT", 91 | "link": 230, 92 | "slot_index": 4 93 | } 94 | ], 95 | "outputs": [ 96 | { 97 | "name": "output", 98 | "type": "LATENT", 99 | "links": [ 100 | 210 101 | ], 102 | "slot_index": 0, 103 | "shape": 3 104 | }, 105 | { 106 | "name": "denoised_output", 107 | "type": "LATENT", 108 | "links": null, 109 | "shape": 3 110 | } 111 | ], 112 | "properties": { 113 | "Node name for S&R": "SamplerCustomAdvanced" 114 | }, 115 | "widgets_values": [] 116 | }, 117 | { 118 | "id": 79, 119 | "type": "DisableNoise", 120 | "pos": [ 121 | 1169.6798095703125, 122 | 124.1329345703125 123 | ], 124 | "size": [ 125 | 210, 126 | 26 127 | ], 128 | "flags": { 129 | "collapsed": true 130 | }, 131 | "order": 0, 132 | "mode": 0, 133 | "inputs": [], 134 | "outputs": [ 135 | { 136 | "name": "NOISE", 137 | "type": "NOISE", 138 | "links": [ 139 | 217 140 | ], 141 | "slot_index": 0 142 | } 143 | ], 144 | "properties": { 145 | "Node name for S&R": "DisableNoise" 146 | }, 147 | "widgets_values": [] 148 | }, 149 | { 150 | "id": 44, 151 | "type": "CLIPTextEncode", 152 | "pos": [ 153 | 420, 154 | 200 155 | ], 156 | "size": [ 157 | 422.84503173828125, 158 | 164.31304931640625 159 | ], 160 | "flags": {}, 161 | "order": 6, 162 | "mode": 0, 163 | "inputs": [ 164 | { 165 | "name": "clip", 166 | "type": "CLIP", 167 | "link": 205 168 | } 169 | ], 170 | "outputs": [ 171 | { 172 | "name": "CONDITIONING", 173 | "type": "CONDITIONING", 174 | "links": [ 175 | 175 176 | ], 177 | "slot_index": 0 178 | } 179 | ], 180 | "title": "CLIP Text Encode (Positive Prompt)", 181 | "properties": { 182 | "Node name for S&R": "CLIPTextEncode" 183 | }, 184 | "widgets_values": [ 185 | "a hippo walking across concrete at the zoo" 186 | ], 187 | "color": "#232", 188 | "bgcolor": "#353" 189 | }, 190 | { 191 | "id": 83, 192 | "type": "CLIPTextEncode", 193 | "pos": [ 194 | 417.337890625, 195 | 411.0529479980469 196 | ], 197 | "size": [ 198 | 422.84503173828125, 199 | 164.31304931640625 200 | ], 201 | "flags": {}, 202 | "order": 7, 203 | "mode": 0, 204 | "inputs": [ 205 | { 206 | "name": "clip", 207 | "type": "CLIP", 208 | "link": 223 209 | } 210 | ], 211 | "outputs": [ 212 | { 213 | "name": "CONDITIONING", 214 | "type": "CONDITIONING", 215 | "links": [ 216 | 224 217 | ], 218 | "slot_index": 0 219 | } 220 | ], 221 | "title": "CLIP Text Encode (Positive Prompt)", 222 | "properties": { 223 | "Node name for S&R": "CLIPTextEncode" 224 | }, 225 | "widgets_values": [ 226 | "a dragon walking across concrete at the zoo" 227 | ], 228 | "color": "#232", 229 | "bgcolor": "#353" 230 | }, 231 | { 232 | "id": 10, 233 | "type": "VAELoader", 234 | "pos": [ 235 | -240, 236 | 340 237 | ], 238 | "size": [ 239 | 210, 240 | 58 241 | ], 242 | "flags": {}, 243 | "order": 1, 244 | "mode": 0, 245 | "inputs": [], 246 | "outputs": [ 247 | { 248 | "name": "VAE", 249 | "type": "VAE", 250 | "links": [ 251 | 211, 252 | 229 253 | ], 254 | "slot_index": 0, 255 | "shape": 3 256 | } 257 | ], 258 | "properties": { 259 | "Node name for S&R": "VAELoader" 260 | }, 261 | "widgets_values": [ 262 | "hunyuan_video_vae_bf16.safetensors" 263 | ] 264 | }, 265 | { 266 | "id": 11, 267 | "type": "DualCLIPLoader", 268 | "pos": [ 269 | -240, 270 | 170 271 | ], 272 | "size": [ 273 | 210, 274 | 106 275 | ], 276 | "flags": {}, 277 | "order": 2, 278 | "mode": 0, 279 | "inputs": [], 280 | "outputs": [ 281 | { 282 | "name": "CLIP", 283 | "type": "CLIP", 284 | "links": [ 285 | 205, 286 | 223 287 | ], 288 | "slot_index": 0, 289 | "shape": 3 290 | } 291 | ], 292 | "properties": { 293 | "Node name for S&R": "DualCLIPLoader" 294 | }, 295 | "widgets_values": [ 296 | "clip_l.safetensors", 297 | "llava_llama3_fp8_scaled.safetensors", 298 | "hunyuan_video" 299 | ] 300 | }, 301 | { 302 | "id": 12, 303 | "type": "UNETLoader", 304 | "pos": [ 305 | -250, 306 | 30 307 | ], 308 | "size": [ 309 | 210, 310 | 82 311 | ], 312 | "flags": {}, 313 | "order": 3, 314 | "mode": 0, 315 | "inputs": [], 316 | "outputs": [ 317 | { 318 | "name": "MODEL", 319 | "type": "MODEL", 320 | "links": [ 321 | 190, 322 | 221 323 | ], 324 | "slot_index": 0, 325 | "shape": 3 326 | } 327 | ], 328 | "properties": { 329 | "Node name for S&R": "UNETLoader" 330 | }, 331 | "widgets_values": [ 332 | "hunyuan_video_t2v_720p_bf16.safetensors", 333 | "fp8_e4m3fn" 334 | ], 335 | "color": "#223", 336 | "bgcolor": "#335" 337 | }, 338 | { 339 | "id": 87, 340 | "type": "VAEEncode", 341 | "pos": [ 342 | 605.8049926757812, 343 | 661.6065063476562 344 | ], 345 | "size": [ 346 | 210, 347 | 46 348 | ], 349 | "flags": {}, 350 | "order": 13, 351 | "mode": 0, 352 | "inputs": [ 353 | { 354 | "name": "pixels", 355 | "type": "IMAGE", 356 | "link": 228 357 | }, 358 | { 359 | "name": "vae", 360 | "type": "VAE", 361 | "link": 229 362 | } 363 | ], 364 | "outputs": [ 365 | { 366 | "name": "LATENT", 367 | "type": "LATENT", 368 | "links": [ 369 | 230 370 | ], 371 | "slot_index": 0 372 | } 373 | ], 374 | "properties": { 375 | "Node name for S&R": "VAEEncode" 376 | }, 377 | "widgets_values": [] 378 | }, 379 | { 380 | "id": 78, 381 | "type": "VHS_VideoCombine", 382 | "pos": [ 383 | 1724.373046875, 384 | 172.32327270507812 385 | ], 386 | "size": [ 387 | 950.9851684570312, 388 | 1046.2388916015625 389 | ], 390 | "flags": {}, 391 | "order": 17, 392 | "mode": 0, 393 | "inputs": [ 394 | { 395 | "name": "images", 396 | "type": "IMAGE", 397 | "link": 216 398 | }, 399 | { 400 | "name": "audio", 401 | "type": "AUDIO", 402 | "link": null, 403 | "shape": 7 404 | }, 405 | { 406 | "name": "meta_batch", 407 | "type": "VHS_BatchManager", 408 | "link": null, 409 | "shape": 7 410 | }, 411 | { 412 | "name": "vae", 413 | "type": "VAE", 414 | "link": null, 415 | "shape": 7 416 | } 417 | ], 418 | "outputs": [ 419 | { 420 | "name": "Filenames", 421 | "type": "VHS_FILENAMES", 422 | "links": null 423 | } 424 | ], 425 | "properties": { 426 | "Node name for S&R": "VHS_VideoCombine" 427 | }, 428 | "widgets_values": { 429 | "frame_rate": 24, 430 | "loop_count": 0, 431 | "filename_prefix": "hyloom", 432 | "format": "video/h264-mp4", 433 | "pix_fmt": "yuv420p", 434 | "crf": 19, 435 | "save_metadata": false, 436 | "trim_to_audio": false, 437 | "pingpong": false, 438 | "save_output": false, 439 | "videopreview": { 440 | "hidden": false, 441 | "paused": false, 442 | "params": { 443 | "filename": "hyloom_00007.mp4", 444 | "subfolder": "", 445 | "type": "temp", 446 | "format": "video/h264-mp4", 447 | "frame_rate": 24, 448 | "workflow": "hyloom_00007.png", 449 | "fullpath": "/workspace/ComfyUI/temp/hyloom_00007.mp4" 450 | }, 451 | "muted": false 452 | } 453 | } 454 | }, 455 | { 456 | "id": 85, 457 | "type": "VHS_LoadVideo", 458 | "pos": [ 459 | -280, 460 | 470 461 | ], 462 | "size": [ 463 | 252.056640625, 464 | 262 465 | ], 466 | "flags": {}, 467 | "order": 4, 468 | "mode": 0, 469 | "inputs": [ 470 | { 471 | "name": "meta_batch", 472 | "type": "VHS_BatchManager", 473 | "link": null, 474 | "shape": 7 475 | }, 476 | { 477 | "name": "vae", 478 | "type": "VAE", 479 | "link": null, 480 | "shape": 7 481 | } 482 | ], 483 | "outputs": [ 484 | { 485 | "name": "IMAGE", 486 | "type": "IMAGE", 487 | "links": [ 488 | 227 489 | ], 490 | "slot_index": 0 491 | }, 492 | { 493 | "name": "frame_count", 494 | "type": "INT", 495 | "links": null 496 | }, 497 | { 498 | "name": "audio", 499 | "type": "AUDIO", 500 | "links": null 501 | }, 502 | { 503 | "name": "video_info", 504 | "type": "VHS_VIDEOINFO", 505 | "links": null 506 | } 507 | ], 508 | "properties": { 509 | "Node name for S&R": "VHS_LoadVideo" 510 | }, 511 | "widgets_values": { 512 | "video": "15448669-hd_1920_1080_60fps (1).mp4", 513 | "force_rate": 0, 514 | "force_size": "Disabled", 515 | "custom_width": 512, 516 | "custom_height": 512, 517 | "frame_load_cap": 49, 518 | "skip_first_frames": 0, 519 | "select_every_nth": 6, 520 | "choose video to upload": "image", 521 | "videopreview": { 522 | "hidden": false, 523 | "paused": false, 524 | "params": { 525 | "force_rate": 0, 526 | "frame_load_cap": 49, 527 | "skip_first_frames": 0, 528 | "select_every_nth": 6, 529 | "filename": "15448669-hd_1920_1080_60fps (1).mp4", 530 | "type": "input", 531 | "format": "video/mp4" 532 | }, 533 | "muted": false 534 | } 535 | } 536 | }, 537 | { 538 | "id": 86, 539 | "type": "ImageScale", 540 | "pos": [ 541 | 9.297161102294922, 542 | 476.7259521484375 543 | ], 544 | "size": [ 545 | 210, 546 | 130 547 | ], 548 | "flags": {}, 549 | "order": 10, 550 | "mode": 0, 551 | "inputs": [ 552 | { 553 | "name": "image", 554 | "type": "IMAGE", 555 | "link": 227 556 | } 557 | ], 558 | "outputs": [ 559 | { 560 | "name": "IMAGE", 561 | "type": "IMAGE", 562 | "links": [ 563 | 228 564 | ], 565 | "slot_index": 0 566 | } 567 | ], 568 | "properties": { 569 | "Node name for S&R": "ImageScale" 570 | }, 571 | "widgets_values": [ 572 | "nearest-exact", 573 | 640, 574 | 480, 575 | "center" 576 | ] 577 | }, 578 | { 579 | "id": 26, 580 | "type": "FluxGuidance", 581 | "pos": [ 582 | 852.5541381835938, 583 | 192.33192443847656 584 | ], 585 | "size": [ 586 | 211.60000610351562, 587 | 58 588 | ], 589 | "flags": {}, 590 | "order": 11, 591 | "mode": 0, 592 | "inputs": [ 593 | { 594 | "name": "conditioning", 595 | "type": "CONDITIONING", 596 | "link": 175 597 | } 598 | ], 599 | "outputs": [ 600 | { 601 | "name": "CONDITIONING", 602 | "type": "CONDITIONING", 603 | "links": [ 604 | 225 605 | ], 606 | "slot_index": 0, 607 | "shape": 3 608 | } 609 | ], 610 | "properties": { 611 | "Node name for S&R": "FluxGuidance" 612 | }, 613 | "widgets_values": [ 614 | 1 615 | ], 616 | "color": "#233", 617 | "bgcolor": "#355" 618 | }, 619 | { 620 | "id": 84, 621 | "type": "FluxGuidance", 622 | "pos": [ 623 | 853.2695922851562, 624 | 425.6966857910156 625 | ], 626 | "size": [ 627 | 211.60000610351562, 628 | 58 629 | ], 630 | "flags": {}, 631 | "order": 12, 632 | "mode": 0, 633 | "inputs": [ 634 | { 635 | "name": "conditioning", 636 | "type": "CONDITIONING", 637 | "link": 224 638 | } 639 | ], 640 | "outputs": [ 641 | { 642 | "name": "CONDITIONING", 643 | "type": "CONDITIONING", 644 | "links": [ 645 | 226 646 | ], 647 | "slot_index": 0, 648 | "shape": 3 649 | } 650 | ], 651 | "properties": { 652 | "Node name for S&R": "FluxGuidance" 653 | }, 654 | "widgets_values": [ 655 | 7 656 | ], 657 | "color": "#233", 658 | "bgcolor": "#355" 659 | }, 660 | { 661 | "id": 81, 662 | "type": "HYFlowEditGuider", 663 | "pos": [ 664 | 1173.615478515625, 665 | 16.504831314086914 666 | ], 667 | "size": [ 668 | 203.2571258544922, 669 | 69.3235855102539 670 | ], 671 | "flags": {}, 672 | "order": 14, 673 | "mode": 0, 674 | "inputs": [ 675 | { 676 | "name": "model", 677 | "type": "MODEL", 678 | "link": 222 679 | }, 680 | { 681 | "name": "source_cond", 682 | "type": "CONDITIONING", 683 | "link": 225 684 | }, 685 | { 686 | "name": "target_cond", 687 | "type": "CONDITIONING", 688 | "link": 226 689 | } 690 | ], 691 | "outputs": [ 692 | { 693 | "name": "GUIDER", 694 | "type": "GUIDER", 695 | "links": [ 696 | 218 697 | ], 698 | "slot_index": 0 699 | } 700 | ], 701 | "properties": { 702 | "Node name for S&R": "HYFlowEditGuider" 703 | }, 704 | "widgets_values": [] 705 | }, 706 | { 707 | "id": 89, 708 | "type": "HYFlowEditSampler", 709 | "pos": [ 710 | 1160.8524169921875, 711 | 349.7723693847656 712 | ], 713 | "size": [ 714 | 219.18406677246094, 715 | 130 716 | ], 717 | "flags": {}, 718 | "order": 5, 719 | "mode": 0, 720 | "inputs": [], 721 | "outputs": [ 722 | { 723 | "name": "SAMPLER", 724 | "type": "SAMPLER", 725 | "links": [ 726 | 231 727 | ], 728 | "slot_index": 0 729 | } 730 | ], 731 | "properties": { 732 | "Node name for S&R": "HYFlowEditSampler" 733 | }, 734 | "widgets_values": [ 735 | 2, 736 | 8, 737 | 0, 738 | "fixed" 739 | ] 740 | }, 741 | { 742 | "id": 17, 743 | "type": "BasicScheduler", 744 | "pos": [ 745 | 1162.681640625, 746 | 528.71142578125 747 | ], 748 | "size": [ 749 | 210, 750 | 106 751 | ], 752 | "flags": {}, 753 | "order": 8, 754 | "mode": 0, 755 | "inputs": [ 756 | { 757 | "name": "model", 758 | "type": "MODEL", 759 | "link": 190, 760 | "slot_index": 0 761 | } 762 | ], 763 | "outputs": [ 764 | { 765 | "name": "SIGMAS", 766 | "type": "SIGMAS", 767 | "links": [ 768 | 20 769 | ], 770 | "shape": 3 771 | } 772 | ], 773 | "properties": { 774 | "Node name for S&R": "BasicScheduler" 775 | }, 776 | "widgets_values": [ 777 | "simple", 778 | 30, 779 | 1 780 | ] 781 | }, 782 | { 783 | "id": 82, 784 | "type": "HYReverseModelSamplingPred", 785 | "pos": [ 786 | 448.6692199707031, 787 | 58.04368209838867 788 | ], 789 | "size": [ 790 | 218.39999389648438, 791 | 58 792 | ], 793 | "flags": {}, 794 | "order": 9, 795 | "mode": 0, 796 | "inputs": [ 797 | { 798 | "name": "model", 799 | "type": "MODEL", 800 | "link": 221 801 | } 802 | ], 803 | "outputs": [ 804 | { 805 | "name": "MODEL", 806 | "type": "MODEL", 807 | "links": [ 808 | 222 809 | ], 810 | "slot_index": 0 811 | } 812 | ], 813 | "properties": { 814 | "Node name for S&R": "HYReverseModelSamplingPred" 815 | }, 816 | "widgets_values": [ 817 | 7 818 | ] 819 | } 820 | ], 821 | "links": [ 822 | [ 823 | 20, 824 | 17, 825 | 0, 826 | 13, 827 | 3, 828 | "SIGMAS" 829 | ], 830 | [ 831 | 175, 832 | 44, 833 | 0, 834 | 26, 835 | 0, 836 | "CONDITIONING" 837 | ], 838 | [ 839 | 190, 840 | 12, 841 | 0, 842 | 17, 843 | 0, 844 | "MODEL" 845 | ], 846 | [ 847 | 205, 848 | 11, 849 | 0, 850 | 44, 851 | 0, 852 | "CLIP" 853 | ], 854 | [ 855 | 210, 856 | 13, 857 | 0, 858 | 73, 859 | 0, 860 | "LATENT" 861 | ], 862 | [ 863 | 211, 864 | 10, 865 | 0, 866 | 73, 867 | 1, 868 | "VAE" 869 | ], 870 | [ 871 | 216, 872 | 73, 873 | 0, 874 | 78, 875 | 0, 876 | "IMAGE" 877 | ], 878 | [ 879 | 217, 880 | 79, 881 | 0, 882 | 13, 883 | 0, 884 | "NOISE" 885 | ], 886 | [ 887 | 218, 888 | 81, 889 | 0, 890 | 13, 891 | 1, 892 | "GUIDER" 893 | ], 894 | [ 895 | 221, 896 | 12, 897 | 0, 898 | 82, 899 | 0, 900 | "MODEL" 901 | ], 902 | [ 903 | 222, 904 | 82, 905 | 0, 906 | 81, 907 | 0, 908 | "MODEL" 909 | ], 910 | [ 911 | 223, 912 | 11, 913 | 0, 914 | 83, 915 | 0, 916 | "CLIP" 917 | ], 918 | [ 919 | 224, 920 | 83, 921 | 0, 922 | 84, 923 | 0, 924 | "CONDITIONING" 925 | ], 926 | [ 927 | 225, 928 | 26, 929 | 0, 930 | 81, 931 | 1, 932 | "CONDITIONING" 933 | ], 934 | [ 935 | 226, 936 | 84, 937 | 0, 938 | 81, 939 | 2, 940 | "CONDITIONING" 941 | ], 942 | [ 943 | 227, 944 | 85, 945 | 0, 946 | 86, 947 | 0, 948 | "IMAGE" 949 | ], 950 | [ 951 | 228, 952 | 86, 953 | 0, 954 | 87, 955 | 0, 956 | "IMAGE" 957 | ], 958 | [ 959 | 229, 960 | 10, 961 | 0, 962 | 87, 963 | 1, 964 | "VAE" 965 | ], 966 | [ 967 | 230, 968 | 87, 969 | 0, 970 | 13, 971 | 4, 972 | "LATENT" 973 | ], 974 | [ 975 | 231, 976 | 89, 977 | 0, 978 | 13, 979 | 2, 980 | "SAMPLER" 981 | ] 982 | ], 983 | "groups": [], 984 | "config": {}, 985 | "extra": { 986 | "ds": { 987 | "scale": 0.391425130122103, 988 | "offset": [ 989 | 952.7087289433513, 990 | 499.0369677684153 991 | ] 992 | }, 993 | "groupNodes": {} 994 | }, 995 | "version": 0.4 996 | } -------------------------------------------------------------------------------- /example_workflows/wrapper_example_hunyuan_flowedit.json: -------------------------------------------------------------------------------- 1 | { 2 | "last_node_id": 86, 3 | "last_link_id": 167, 4 | "nodes": [ 5 | { 6 | "id": 65, 7 | "type": "HyVideoTextEncode", 8 | "pos": [ 9 | 569.6162109375, 10 | -855.2640991210938 11 | ], 12 | "size": [ 13 | 459.8385009765625, 14 | 345.2080383300781 15 | ], 16 | "flags": {}, 17 | "order": 6, 18 | "mode": 0, 19 | "inputs": [ 20 | { 21 | "name": "text_encoders", 22 | "type": "HYVIDTEXTENCODER", 23 | "link": 102 24 | }, 25 | { 26 | "name": "custom_prompt_template", 27 | "type": "PROMPT_TEMPLATE", 28 | "link": null, 29 | "shape": 7 30 | }, 31 | { 32 | "name": "clip_l", 33 | "type": "CLIP", 34 | "link": 103, 35 | "shape": 7 36 | }, 37 | { 38 | "name": "hyvid_cfg", 39 | "type": "HYVID_CFG", 40 | "link": null, 41 | "shape": 7 42 | } 43 | ], 44 | "outputs": [ 45 | { 46 | "name": "hyvid_embeds", 47 | "type": "HYVIDEMBEDS", 48 | "links": [ 49 | 164 50 | ], 51 | "slot_index": 0 52 | } 53 | ], 54 | "properties": { 55 | "Node name for S&R": "HyVideoTextEncode" 56 | }, 57 | "widgets_values": [ 58 | "Main content and theme: cinematic and photorealistic video of a helicopter hovering in place while resting on the platform of a helipad \nObject details: \na helicopter \nRainforest background \nThe video style is very cinematic and has dramatic lighting \nActions and movements: a helicopter's rotor blades are spinning \nBackground and atmosphere: \nThe background is a peaceful rainforest, which creates a serene, calm atmosphere \nmasterpiece, best quality ", 59 | true, 60 | "video" 61 | ] 62 | }, 63 | { 64 | "id": 85, 65 | "type": "HyVideoTextEncode", 66 | "pos": [ 67 | 572.2459716796875, 68 | -1254.3387451171875 69 | ], 70 | "size": [ 71 | 459.8385009765625, 72 | 345.2080383300781 73 | ], 74 | "flags": {}, 75 | "order": 7, 76 | "mode": 0, 77 | "inputs": [ 78 | { 79 | "name": "text_encoders", 80 | "type": "HYVIDTEXTENCODER", 81 | "link": 158 82 | }, 83 | { 84 | "name": "custom_prompt_template", 85 | "type": "PROMPT_TEMPLATE", 86 | "link": null, 87 | "shape": 7 88 | }, 89 | { 90 | "name": "clip_l", 91 | "type": "CLIP", 92 | "link": 157, 93 | "shape": 7 94 | }, 95 | { 96 | "name": "hyvid_cfg", 97 | "type": "HYVID_CFG", 98 | "link": null, 99 | "shape": 7 100 | } 101 | ], 102 | "outputs": [ 103 | { 104 | "name": "hyvid_embeds", 105 | "type": "HYVIDEMBEDS", 106 | "links": [ 107 | 163 108 | ], 109 | "slot_index": 0 110 | } 111 | ], 112 | "properties": { 113 | "Node name for S&R": "HyVideoTextEncode" 114 | }, 115 | "widgets_values": [ 116 | "Main content and theme: cinematic and photorealistic video of a parrot flapping its wings while perched on the hand of a woman\nObject details:\na parrot\nRainforest background\nThe video style is very cinematic and has dramatic lighting\nActions and movements: a parrot is flapping its wings\nBackground and atmosphere:\nThe background is a peaceful rainforest, which creates a serene, calm atmosphere\nmasterpiece, best quality", 117 | true, 118 | "video" 119 | ] 120 | }, 121 | { 122 | "id": 71, 123 | "type": "GetImageSizeAndCount", 124 | "pos": [ 125 | 438.6930236816406, 126 | -58.08594512939453 127 | ], 128 | "size": [ 129 | 277.20001220703125, 130 | 86 131 | ], 132 | "flags": {}, 133 | "order": 10, 134 | "mode": 0, 135 | "inputs": [ 136 | { 137 | "name": "image", 138 | "type": "IMAGE", 139 | "link": 120 140 | } 141 | ], 142 | "outputs": [ 143 | { 144 | "name": "image", 145 | "type": "IMAGE", 146 | "links": [ 147 | 117 148 | ], 149 | "slot_index": 0 150 | }, 151 | { 152 | "name": "960 width", 153 | "type": "INT", 154 | "links": [], 155 | "slot_index": 1 156 | }, 157 | { 158 | "name": "544 height", 159 | "type": "INT", 160 | "links": [], 161 | "slot_index": 2 162 | }, 163 | { 164 | "name": "49 count", 165 | "type": "INT", 166 | "links": [], 167 | "slot_index": 3 168 | } 169 | ], 170 | "properties": { 171 | "Node name for S&R": "GetImageSizeAndCount" 172 | }, 173 | "widgets_values": [] 174 | }, 175 | { 176 | "id": 72, 177 | "type": "SetNode", 178 | "pos": [ 179 | 449.2713317871094, 180 | 97.60063934326172 181 | ], 182 | "size": [ 183 | 210, 184 | 58 185 | ], 186 | "flags": { 187 | "collapsed": true 188 | }, 189 | "order": 9, 190 | "mode": 0, 191 | "inputs": [ 192 | { 193 | "name": "IMAGE", 194 | "type": "IMAGE", 195 | "link": 119 196 | } 197 | ], 198 | "outputs": [ 199 | { 200 | "name": "IMAGE", 201 | "type": "IMAGE", 202 | "links": [ 203 | 120 204 | ], 205 | "slot_index": 0 206 | } 207 | ], 208 | "title": "Set_InputVideo", 209 | "properties": { 210 | "previousName": "InputVideo" 211 | }, 212 | "widgets_values": [ 213 | "InputVideo" 214 | ], 215 | "color": "#2a363b", 216 | "bgcolor": "#3f5159" 217 | }, 218 | { 219 | "id": 62, 220 | "type": "ImageResizeKJ", 221 | "pos": [ 222 | 96.65345764160156, 223 | -54.70817947387695 224 | ], 225 | "size": [ 226 | 315, 227 | 266 228 | ], 229 | "flags": {}, 230 | "order": 8, 231 | "mode": 0, 232 | "inputs": [ 233 | { 234 | "name": "image", 235 | "type": "IMAGE", 236 | "link": 75 237 | }, 238 | { 239 | "name": "get_image_size", 240 | "type": "IMAGE", 241 | "link": null, 242 | "shape": 7 243 | }, 244 | { 245 | "name": "width_input", 246 | "type": "INT", 247 | "link": null, 248 | "widget": { 249 | "name": "width_input" 250 | }, 251 | "shape": 7 252 | }, 253 | { 254 | "name": "height_input", 255 | "type": "INT", 256 | "link": null, 257 | "widget": { 258 | "name": "height_input" 259 | }, 260 | "shape": 7 261 | } 262 | ], 263 | "outputs": [ 264 | { 265 | "name": "IMAGE", 266 | "type": "IMAGE", 267 | "links": [ 268 | 119 269 | ], 270 | "slot_index": 0 271 | }, 272 | { 273 | "name": "width", 274 | "type": "INT", 275 | "links": [], 276 | "slot_index": 1 277 | }, 278 | { 279 | "name": "height", 280 | "type": "INT", 281 | "links": [], 282 | "slot_index": 2 283 | } 284 | ], 285 | "properties": { 286 | "Node name for S&R": "ImageResizeKJ" 287 | }, 288 | "widgets_values": [ 289 | 960, 290 | 544, 291 | "lanczos", 292 | false, 293 | 16, 294 | 0, 295 | 0, 296 | "center" 297 | ] 298 | }, 299 | { 300 | "id": 16, 301 | "type": "DownloadAndLoadHyVideoTextEncoder", 302 | "pos": [ 303 | -310.33868408203125, 304 | -1085.5028076171875 305 | ], 306 | "size": [ 307 | 429.6711730957031, 308 | 178 309 | ], 310 | "flags": {}, 311 | "order": 0, 312 | "mode": 0, 313 | "inputs": [], 314 | "outputs": [ 315 | { 316 | "name": "hyvid_text_encoder", 317 | "type": "HYVIDTEXTENCODER", 318 | "links": [ 319 | 102, 320 | 158 321 | ], 322 | "slot_index": 0 323 | } 324 | ], 325 | "properties": { 326 | "Node name for S&R": "DownloadAndLoadHyVideoTextEncoder" 327 | }, 328 | "widgets_values": [ 329 | "Kijai/llava-llama-3-8b-text-encoder-tokenizer", 330 | "disabled", 331 | "fp16", 332 | false, 333 | 2, 334 | "disabled" 335 | ] 336 | }, 337 | { 338 | "id": 43, 339 | "type": "CLIPLoader", 340 | "pos": [ 341 | -190.31114196777344, 342 | -834.338134765625 343 | ], 344 | "size": [ 345 | 267.4123840332031, 346 | 82 347 | ], 348 | "flags": {}, 349 | "order": 1, 350 | "mode": 0, 351 | "inputs": [], 352 | "outputs": [ 353 | { 354 | "name": "CLIP", 355 | "type": "CLIP", 356 | "links": [ 357 | 103, 358 | 157 359 | ], 360 | "slot_index": 0 361 | } 362 | ], 363 | "properties": { 364 | "Node name for S&R": "CLIPLoader" 365 | }, 366 | "widgets_values": [ 367 | "clip_l.safetensors", 368 | "stable_diffusion", 369 | "default" 370 | ] 371 | }, 372 | { 373 | "id": 86, 374 | "type": "HyVideoFlowEditSamplerWrapper", 375 | "pos": [ 376 | 1153.6915283203125, 377 | -219.83078002929688 378 | ], 379 | "size": [ 380 | 400.9231262207031, 381 | 570 382 | ], 383 | "flags": {}, 384 | "order": 12, 385 | "mode": 0, 386 | "inputs": [ 387 | { 388 | "name": "model", 389 | "type": "HYVIDEOMODEL", 390 | "link": 167 391 | }, 392 | { 393 | "name": "source_embeds", 394 | "type": "HYVIDEMBEDS", 395 | "link": 163 396 | }, 397 | { 398 | "name": "target_embeds", 399 | "type": "HYVIDEMBEDS", 400 | "link": 164 401 | }, 402 | { 403 | "name": "samples", 404 | "type": "LATENT", 405 | "link": 165 406 | } 407 | ], 408 | "outputs": [ 409 | { 410 | "name": "samples", 411 | "type": "LATENT", 412 | "links": [ 413 | 166 414 | ] 415 | } 416 | ], 417 | "properties": { 418 | "Node name for S&R": "HyVideoFlowEditSamplerWrapper" 419 | }, 420 | "widgets_values": [ 421 | 30, 422 | 4, 423 | 0, 424 | 6, 425 | 12, 426 | 6, 427 | 6, 428 | 3, 429 | 0, 430 | "fixed", 431 | true 432 | ] 433 | }, 434 | { 435 | "id": 7, 436 | "type": "HyVideoVAELoader", 437 | "pos": [ 438 | -399.6872253417969, 439 | -457.69366455078125 440 | ], 441 | "size": [ 442 | 466.5, 443 | 82 444 | ], 445 | "flags": {}, 446 | "order": 2, 447 | "mode": 0, 448 | "inputs": [ 449 | { 450 | "name": "compile_args", 451 | "type": "COMPILEARGS", 452 | "link": null, 453 | "shape": 7 454 | } 455 | ], 456 | "outputs": [ 457 | { 458 | "name": "vae", 459 | "type": "VAE", 460 | "links": [ 461 | 73, 462 | 105 463 | ], 464 | "slot_index": 0 465 | } 466 | ], 467 | "properties": { 468 | "Node name for S&R": "HyVideoVAELoader" 469 | }, 470 | "widgets_values": [ 471 | "hyvid\\hunyuan_video_vae_bf16.safetensors", 472 | "bf16" 473 | ] 474 | }, 475 | { 476 | "id": 1, 477 | "type": "HyVideoModelLoader", 478 | "pos": [ 479 | -431.7147521972656, 480 | -310.651123046875 481 | ], 482 | "size": [ 483 | 509.7506103515625, 484 | 218 485 | ], 486 | "flags": {}, 487 | "order": 3, 488 | "mode": 0, 489 | "inputs": [ 490 | { 491 | "name": "compile_args", 492 | "type": "COMPILEARGS", 493 | "link": null, 494 | "shape": 7 495 | }, 496 | { 497 | "name": "block_swap_args", 498 | "type": "BLOCKSWAPARGS", 499 | "link": null, 500 | "shape": 7 501 | }, 502 | { 503 | "name": "lora", 504 | "type": "HYVIDLORA", 505 | "link": null, 506 | "shape": 7 507 | } 508 | ], 509 | "outputs": [ 510 | { 511 | "name": "model", 512 | "type": "HYVIDEOMODEL", 513 | "links": [ 514 | 167 515 | ], 516 | "slot_index": 0 517 | } 518 | ], 519 | "properties": { 520 | "Node name for S&R": "HyVideoModelLoader" 521 | }, 522 | "widgets_values": [ 523 | "hyvideo\\hunyuan_video_720_fp8_e4m3fn.safetensors", 524 | "bf16", 525 | "fp8_e4m3fn", 526 | "offload_device", 527 | "sageattn_varlen", 528 | false 529 | ] 530 | }, 531 | { 532 | "id": 60, 533 | "type": "VHS_LoadVideo", 534 | "pos": [ 535 | -215.327880859375, 536 | -22.302289962768555 537 | ], 538 | "size": [ 539 | 252.056640625, 540 | 508.056640625 541 | ], 542 | "flags": {}, 543 | "order": 4, 544 | "mode": 0, 545 | "inputs": [ 546 | { 547 | "name": "meta_batch", 548 | "type": "VHS_BatchManager", 549 | "link": null, 550 | "shape": 7 551 | }, 552 | { 553 | "name": "vae", 554 | "type": "VAE", 555 | "link": null, 556 | "shape": 7 557 | } 558 | ], 559 | "outputs": [ 560 | { 561 | "name": "IMAGE", 562 | "type": "IMAGE", 563 | "links": [ 564 | 75 565 | ], 566 | "slot_index": 0 567 | }, 568 | { 569 | "name": "frame_count", 570 | "type": "INT", 571 | "links": [], 572 | "slot_index": 1 573 | }, 574 | { 575 | "name": "audio", 576 | "type": "AUDIO", 577 | "links": null 578 | }, 579 | { 580 | "name": "video_info", 581 | "type": "VHS_VIDEOINFO", 582 | "links": null 583 | } 584 | ], 585 | "properties": { 586 | "Node name for S&R": "VHS_LoadVideo" 587 | }, 588 | "widgets_values": { 589 | "video": "wolf_interpolated.mp4", 590 | "force_rate": 0, 591 | "force_size": "Disabled", 592 | "custom_width": 512, 593 | "custom_height": 512, 594 | "frame_load_cap": 49, 595 | "skip_first_frames": 0, 596 | "select_every_nth": 1, 597 | "choose video to upload": "image", 598 | "videopreview": { 599 | "hidden": false, 600 | "paused": false, 601 | "params": { 602 | "frame_load_cap": 49, 603 | "skip_first_frames": 0, 604 | "force_rate": 0, 605 | "filename": "wolf_interpolated.mp4", 606 | "type": "input", 607 | "format": "video/mp4", 608 | "select_every_nth": 1 609 | }, 610 | "muted": false 611 | } 612 | } 613 | }, 614 | { 615 | "id": 61, 616 | "type": "HyVideoEncode", 617 | "pos": [ 618 | 772.994140625, 619 | -137.0120849609375 620 | ], 621 | "size": [ 622 | 315, 623 | 150 624 | ], 625 | "flags": {}, 626 | "order": 11, 627 | "mode": 0, 628 | "inputs": [ 629 | { 630 | "name": "vae", 631 | "type": "VAE", 632 | "link": 73 633 | }, 634 | { 635 | "name": "image", 636 | "type": "IMAGE", 637 | "link": 117 638 | } 639 | ], 640 | "outputs": [ 641 | { 642 | "name": "samples", 643 | "type": "LATENT", 644 | "links": [ 645 | 165 646 | ], 647 | "slot_index": 0 648 | } 649 | ], 650 | "properties": { 651 | "Node name for S&R": "HyVideoEncode" 652 | }, 653 | "widgets_values": [ 654 | true, 655 | 64, 656 | 256, 657 | true 658 | ] 659 | }, 660 | { 661 | "id": 66, 662 | "type": "HyVideoDecode", 663 | "pos": [ 664 | 1228.44970703125, 665 | -446.55419921875 666 | ], 667 | "size": [ 668 | 292.81866455078125, 669 | 150 670 | ], 671 | "flags": {}, 672 | "order": 13, 673 | "mode": 0, 674 | "inputs": [ 675 | { 676 | "name": "vae", 677 | "type": "VAE", 678 | "link": 105 679 | }, 680 | { 681 | "name": "samples", 682 | "type": "LATENT", 683 | "link": 166 684 | } 685 | ], 686 | "outputs": [ 687 | { 688 | "name": "images", 689 | "type": "IMAGE", 690 | "links": [ 691 | 113 692 | ], 693 | "slot_index": 0 694 | } 695 | ], 696 | "properties": { 697 | "Node name for S&R": "HyVideoDecode" 698 | }, 699 | "widgets_values": [ 700 | true, 701 | 64, 702 | 128, 703 | true 704 | ] 705 | }, 706 | { 707 | "id": 68, 708 | "type": "ImageConcatMulti", 709 | "pos": [ 710 | 1544.142578125, 711 | -461.7588806152344 712 | ], 713 | "size": [ 714 | 210, 715 | 150 716 | ], 717 | "flags": {}, 718 | "order": 14, 719 | "mode": 0, 720 | "inputs": [ 721 | { 722 | "name": "image_1", 723 | "type": "IMAGE", 724 | "link": 121 725 | }, 726 | { 727 | "name": "image_2", 728 | "type": "IMAGE", 729 | "link": 113 730 | } 731 | ], 732 | "outputs": [ 733 | { 734 | "name": "images", 735 | "type": "IMAGE", 736 | "links": [ 737 | 114 738 | ], 739 | "slot_index": 0 740 | } 741 | ], 742 | "properties": {}, 743 | "widgets_values": [ 744 | 2, 745 | "right", 746 | false, 747 | null 748 | ] 749 | }, 750 | { 751 | "id": 73, 752 | "type": "GetNode", 753 | "pos": [ 754 | 1546.0333251953125, 755 | -527.7788696289062 756 | ], 757 | "size": [ 758 | 210, 759 | 58 760 | ], 761 | "flags": { 762 | "collapsed": true 763 | }, 764 | "order": 5, 765 | "mode": 0, 766 | "inputs": [], 767 | "outputs": [ 768 | { 769 | "name": "IMAGE", 770 | "type": "IMAGE", 771 | "links": [ 772 | 121 773 | ], 774 | "slot_index": 0 775 | } 776 | ], 777 | "title": "Get_InputVideo", 778 | "properties": {}, 779 | "widgets_values": [ 780 | "InputVideo" 781 | ], 782 | "color": "#2a363b", 783 | "bgcolor": "#3f5159" 784 | }, 785 | { 786 | "id": 69, 787 | "type": "VHS_VideoCombine", 788 | "pos": [ 789 | 1822.6951904296875, 790 | -464.15716552734375 791 | ], 792 | "size": [ 793 | 1098.79248046875, 794 | 334 795 | ], 796 | "flags": {}, 797 | "order": 15, 798 | "mode": 0, 799 | "inputs": [ 800 | { 801 | "name": "images", 802 | "type": "IMAGE", 803 | "link": 114 804 | }, 805 | { 806 | "name": "audio", 807 | "type": "AUDIO", 808 | "link": null, 809 | "shape": 7 810 | }, 811 | { 812 | "name": "meta_batch", 813 | "type": "VHS_BatchManager", 814 | "link": null, 815 | "shape": 7 816 | }, 817 | { 818 | "name": "vae", 819 | "type": "VAE", 820 | "link": null, 821 | "shape": 7 822 | } 823 | ], 824 | "outputs": [ 825 | { 826 | "name": "Filenames", 827 | "type": "VHS_FILENAMES", 828 | "links": null 829 | } 830 | ], 831 | "properties": { 832 | "Node name for S&R": "VHS_VideoCombine" 833 | }, 834 | "widgets_values": { 835 | "frame_rate": 24, 836 | "loop_count": 0, 837 | "filename_prefix": "hunyuanloom", 838 | "format": "video/h264-mp4", 839 | "pix_fmt": "yuv420p", 840 | "crf": 19, 841 | "save_metadata": false, 842 | "trim_to_audio": false, 843 | "pingpong": false, 844 | "save_output": true, 845 | "videopreview": { 846 | "hidden": false, 847 | "paused": false, 848 | "params": { 849 | "filename": "HunyuanVideo_00014.mp4", 850 | "subfolder": "", 851 | "type": "temp", 852 | "format": "video/h264-mp4", 853 | "frame_rate": 24, 854 | "workflow": "HunyuanVideo_00014.png", 855 | "fullpath": "/workspace/ComfyUI/temp/HunyuanVideo_00014.mp4" 856 | }, 857 | "muted": false 858 | } 859 | } 860 | } 861 | ], 862 | "links": [ 863 | [ 864 | 73, 865 | 7, 866 | 0, 867 | 61, 868 | 0, 869 | "VAE" 870 | ], 871 | [ 872 | 75, 873 | 60, 874 | 0, 875 | 62, 876 | 0, 877 | "IMAGE" 878 | ], 879 | [ 880 | 102, 881 | 16, 882 | 0, 883 | 65, 884 | 0, 885 | "HYVIDTEXTENCODER" 886 | ], 887 | [ 888 | 103, 889 | 43, 890 | 0, 891 | 65, 892 | 2, 893 | "CLIP" 894 | ], 895 | [ 896 | 105, 897 | 7, 898 | 0, 899 | 66, 900 | 0, 901 | "VAE" 902 | ], 903 | [ 904 | 113, 905 | 66, 906 | 0, 907 | 68, 908 | 1, 909 | "IMAGE" 910 | ], 911 | [ 912 | 114, 913 | 68, 914 | 0, 915 | 69, 916 | 0, 917 | "IMAGE" 918 | ], 919 | [ 920 | 117, 921 | 71, 922 | 0, 923 | 61, 924 | 1, 925 | "IMAGE" 926 | ], 927 | [ 928 | 119, 929 | 62, 930 | 0, 931 | 72, 932 | 0, 933 | "*" 934 | ], 935 | [ 936 | 120, 937 | 72, 938 | 0, 939 | 71, 940 | 0, 941 | "IMAGE" 942 | ], 943 | [ 944 | 121, 945 | 73, 946 | 0, 947 | 68, 948 | 0, 949 | "IMAGE" 950 | ], 951 | [ 952 | 157, 953 | 43, 954 | 0, 955 | 85, 956 | 2, 957 | "CLIP" 958 | ], 959 | [ 960 | 158, 961 | 16, 962 | 0, 963 | 85, 964 | 0, 965 | "HYVIDTEXTENCODER" 966 | ], 967 | [ 968 | 163, 969 | 85, 970 | 0, 971 | 86, 972 | 1, 973 | "HYVIDEMBEDS" 974 | ], 975 | [ 976 | 164, 977 | 65, 978 | 0, 979 | 86, 980 | 2, 981 | "HYVIDEMBEDS" 982 | ], 983 | [ 984 | 165, 985 | 61, 986 | 0, 987 | 86, 988 | 3, 989 | "LATENT" 990 | ], 991 | [ 992 | 166, 993 | 86, 994 | 0, 995 | 66, 996 | 1, 997 | "LATENT" 998 | ], 999 | [ 1000 | 167, 1001 | 1, 1002 | 0, 1003 | 86, 1004 | 0, 1005 | "HYVIDEOMODEL" 1006 | ] 1007 | ], 1008 | "groups": [], 1009 | "config": {}, 1010 | "extra": { 1011 | "ds": { 1012 | "scale": 0.6402987863296646, 1013 | "offset": [ 1014 | 485.09283743323897, 1015 | 1208.6430569039285 1016 | ] 1017 | }, 1018 | "VHS_latentpreview": false, 1019 | "VHS_latentpreviewrate": 0 1020 | }, 1021 | "version": 0.4 1022 | } -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /example_workflows/example_hy_regional_prompting_t2v.json: -------------------------------------------------------------------------------- 1 | { 2 | "last_node_id": 101, 3 | "last_link_id": 277, 4 | "nodes": [ 5 | { 6 | "id": 16, 7 | "type": "KSamplerSelect", 8 | "pos": [ 9 | 484, 10 | 751 11 | ], 12 | "size": [ 13 | 315, 14 | 58 15 | ], 16 | "flags": {}, 17 | "order": 0, 18 | "mode": 0, 19 | "inputs": [], 20 | "outputs": [ 21 | { 22 | "name": "SAMPLER", 23 | "type": "SAMPLER", 24 | "links": [ 25 | 19 26 | ], 27 | "shape": 3 28 | } 29 | ], 30 | "properties": { 31 | "Node name for S&R": "KSamplerSelect" 32 | }, 33 | "widgets_values": [ 34 | "euler" 35 | ] 36 | }, 37 | { 38 | "id": 10, 39 | "type": "VAELoader", 40 | "pos": [ 41 | 0, 42 | 420 43 | ], 44 | "size": [ 45 | 350, 46 | 60 47 | ], 48 | "flags": {}, 49 | "order": 1, 50 | "mode": 0, 51 | "inputs": [], 52 | "outputs": [ 53 | { 54 | "name": "VAE", 55 | "type": "VAE", 56 | "links": [ 57 | 211 58 | ], 59 | "slot_index": 0, 60 | "shape": 3 61 | } 62 | ], 63 | "properties": { 64 | "Node name for S&R": "VAELoader" 65 | }, 66 | "widgets_values": [ 67 | "hunyuan_video_vae_bf16.safetensors" 68 | ] 69 | }, 70 | { 71 | "id": 13, 72 | "type": "SamplerCustomAdvanced", 73 | "pos": [ 74 | 1950, 75 | 190 76 | ], 77 | "size": [ 78 | 272.3617858886719, 79 | 124.53733825683594 80 | ], 81 | "flags": {}, 82 | "order": 30, 83 | "mode": 0, 84 | "inputs": [ 85 | { 86 | "name": "noise", 87 | "type": "NOISE", 88 | "link": 37, 89 | "slot_index": 0 90 | }, 91 | { 92 | "name": "guider", 93 | "type": "GUIDER", 94 | "link": 30, 95 | "slot_index": 1 96 | }, 97 | { 98 | "name": "sampler", 99 | "type": "SAMPLER", 100 | "link": 19, 101 | "slot_index": 2 102 | }, 103 | { 104 | "name": "sigmas", 105 | "type": "SIGMAS", 106 | "link": 20, 107 | "slot_index": 3 108 | }, 109 | { 110 | "name": "latent_image", 111 | "type": "LATENT", 112 | "link": 180, 113 | "slot_index": 4 114 | } 115 | ], 116 | "outputs": [ 117 | { 118 | "name": "output", 119 | "type": "LATENT", 120 | "links": [ 121 | 210 122 | ], 123 | "slot_index": 0, 124 | "shape": 3 125 | }, 126 | { 127 | "name": "denoised_output", 128 | "type": "LATENT", 129 | "links": null, 130 | "shape": 3 131 | } 132 | ], 133 | "properties": { 134 | "Node name for S&R": "SamplerCustomAdvanced" 135 | }, 136 | "widgets_values": [] 137 | }, 138 | { 139 | "id": 22, 140 | "type": "BasicGuider", 141 | "pos": [ 142 | 1637.3946533203125, 143 | 94.65210723876953 144 | ], 145 | "size": [ 146 | 222.3482666015625, 147 | 46 148 | ], 149 | "flags": {}, 150 | "order": 29, 151 | "mode": 0, 152 | "inputs": [ 153 | { 154 | "name": "model", 155 | "type": "MODEL", 156 | "link": 224, 157 | "slot_index": 0 158 | }, 159 | { 160 | "name": "conditioning", 161 | "type": "CONDITIONING", 162 | "link": 129, 163 | "slot_index": 1 164 | } 165 | ], 166 | "outputs": [ 167 | { 168 | "name": "GUIDER", 169 | "type": "GUIDER", 170 | "links": [ 171 | 30 172 | ], 173 | "slot_index": 0, 174 | "shape": 3 175 | } 176 | ], 177 | "properties": { 178 | "Node name for S&R": "BasicGuider" 179 | }, 180 | "widgets_values": [] 181 | }, 182 | { 183 | "id": 82, 184 | "type": "ConfigureModifiedHY", 185 | "pos": [ 186 | 385.38446044921875, 187 | -93.34419250488281 188 | ], 189 | "size": [ 190 | 252, 191 | 26 192 | ], 193 | "flags": {}, 194 | "order": 12, 195 | "mode": 0, 196 | "inputs": [ 197 | { 198 | "name": "model", 199 | "type": "MODEL", 200 | "link": 221 201 | } 202 | ], 203 | "outputs": [ 204 | { 205 | "name": "MODEL", 206 | "type": "MODEL", 207 | "links": [ 208 | 222 209 | ], 210 | "slot_index": 0 211 | } 212 | ], 213 | "properties": { 214 | "Node name for S&R": "ConfigureModifiedHY" 215 | }, 216 | "widgets_values": [] 217 | }, 218 | { 219 | "id": 26, 220 | "type": "FluxGuidance", 221 | "pos": [ 222 | 520, 223 | 100 224 | ], 225 | "size": [ 226 | 317.4000244140625, 227 | 58 228 | ], 229 | "flags": {}, 230 | "order": 21, 231 | "mode": 0, 232 | "inputs": [ 233 | { 234 | "name": "conditioning", 235 | "type": "CONDITIONING", 236 | "link": 175 237 | } 238 | ], 239 | "outputs": [ 240 | { 241 | "name": "CONDITIONING", 242 | "type": "CONDITIONING", 243 | "links": [ 244 | 129, 245 | 226 246 | ], 247 | "slot_index": 0, 248 | "shape": 3 249 | } 250 | ], 251 | "properties": { 252 | "Node name for S&R": "FluxGuidance" 253 | }, 254 | "widgets_values": [ 255 | 6 256 | ], 257 | "color": "#233", 258 | "bgcolor": "#355" 259 | }, 260 | { 261 | "id": 12, 262 | "type": "UNETLoader", 263 | "pos": [ 264 | 0, 265 | 150 266 | ], 267 | "size": [ 268 | 350, 269 | 82 270 | ], 271 | "flags": {}, 272 | "order": 2, 273 | "mode": 0, 274 | "inputs": [], 275 | "outputs": [ 276 | { 277 | "name": "MODEL", 278 | "type": "MODEL", 279 | "links": [ 280 | 190, 281 | 221 282 | ], 283 | "slot_index": 0, 284 | "shape": 3 285 | } 286 | ], 287 | "properties": { 288 | "Node name for S&R": "UNETLoader" 289 | }, 290 | "widgets_values": [ 291 | "hunyuan_video_t2v_720p_bf16.safetensors", 292 | "fp8_e4m3fn" 293 | ], 294 | "color": "#223", 295 | "bgcolor": "#335" 296 | }, 297 | { 298 | "id": 73, 299 | "type": "VAEDecodeTiled", 300 | "pos": [ 301 | 2240, 302 | 190 303 | ], 304 | "size": [ 305 | 210, 306 | 102 307 | ], 308 | "flags": {}, 309 | "order": 31, 310 | "mode": 0, 311 | "inputs": [ 312 | { 313 | "name": "samples", 314 | "type": "LATENT", 315 | "link": 210 316 | }, 317 | { 318 | "name": "vae", 319 | "type": "VAE", 320 | "link": 211 321 | } 322 | ], 323 | "outputs": [ 324 | { 325 | "name": "IMAGE", 326 | "type": "IMAGE", 327 | "links": [ 328 | 227 329 | ], 330 | "slot_index": 0 331 | } 332 | ], 333 | "properties": { 334 | "Node name for S&R": "VAEDecodeTiled" 335 | }, 336 | "widgets_values": [ 337 | 256, 338 | 64 339 | ] 340 | }, 341 | { 342 | "id": 11, 343 | "type": "DualCLIPLoader", 344 | "pos": [ 345 | 0, 346 | 270 347 | ], 348 | "size": [ 349 | 350, 350 | 106 351 | ], 352 | "flags": {}, 353 | "order": 3, 354 | "mode": 0, 355 | "inputs": [], 356 | "outputs": [ 357 | { 358 | "name": "CLIP", 359 | "type": "CLIP", 360 | "links": [ 361 | 205, 362 | 225, 363 | 242, 364 | 243 365 | ], 366 | "slot_index": 0, 367 | "shape": 3 368 | } 369 | ], 370 | "properties": { 371 | "Node name for S&R": "DualCLIPLoader" 372 | }, 373 | "widgets_values": [ 374 | "clip_l.safetensors", 375 | "llava_llama3_fp8_scaled.safetensors", 376 | "hunyuan_video" 377 | ] 378 | }, 379 | { 380 | "id": 67, 381 | "type": "ModelSamplingSD3", 382 | "pos": [ 383 | 751.6845703125, 384 | -31.64067268371582 385 | ], 386 | "size": [ 387 | 210, 388 | 58 389 | ], 390 | "flags": {}, 391 | "order": 20, 392 | "mode": 0, 393 | "inputs": [ 394 | { 395 | "name": "model", 396 | "type": "MODEL", 397 | "link": 222 398 | } 399 | ], 400 | "outputs": [ 401 | { 402 | "name": "MODEL", 403 | "type": "MODEL", 404 | "links": [ 405 | 223 406 | ], 407 | "slot_index": 0 408 | } 409 | ], 410 | "properties": { 411 | "Node name for S&R": "ModelSamplingSD3" 412 | }, 413 | "widgets_values": [ 414 | 9 415 | ] 416 | }, 417 | { 418 | "id": 17, 419 | "type": "BasicScheduler", 420 | "pos": [ 421 | 478, 422 | 860 423 | ], 424 | "size": [ 425 | 315, 426 | 106 427 | ], 428 | "flags": {}, 429 | "order": 11, 430 | "mode": 0, 431 | "inputs": [ 432 | { 433 | "name": "model", 434 | "type": "MODEL", 435 | "link": 190, 436 | "slot_index": 0 437 | } 438 | ], 439 | "outputs": [ 440 | { 441 | "name": "SIGMAS", 442 | "type": "SIGMAS", 443 | "links": [ 444 | 20 445 | ], 446 | "shape": 3 447 | } 448 | ], 449 | "properties": { 450 | "Node name for S&R": "BasicScheduler" 451 | }, 452 | "widgets_values": [ 453 | "simple", 454 | 30, 455 | 1 456 | ] 457 | }, 458 | { 459 | "id": 84, 460 | "type": "VHS_VideoCombine", 461 | "pos": [ 462 | 2509.404296875, 463 | 79.10501098632812 464 | ], 465 | "size": [ 466 | 802.333984375, 467 | 790.83056640625 468 | ], 469 | "flags": {}, 470 | "order": 32, 471 | "mode": 0, 472 | "inputs": [ 473 | { 474 | "name": "images", 475 | "type": "IMAGE", 476 | "link": 227 477 | }, 478 | { 479 | "name": "audio", 480 | "type": "AUDIO", 481 | "link": null, 482 | "shape": 7 483 | }, 484 | { 485 | "name": "meta_batch", 486 | "type": "VHS_BatchManager", 487 | "link": null, 488 | "shape": 7 489 | }, 490 | { 491 | "name": "vae", 492 | "type": "VAE", 493 | "link": null, 494 | "shape": 7 495 | } 496 | ], 497 | "outputs": [ 498 | { 499 | "name": "Filenames", 500 | "type": "VHS_FILENAMES", 501 | "links": null 502 | } 503 | ], 504 | "properties": { 505 | "Node name for S&R": "VHS_VideoCombine" 506 | }, 507 | "widgets_values": { 508 | "frame_rate": 24, 509 | "loop_count": 0, 510 | "filename_prefix": "AnimateDiff", 511 | "format": "video/h264-mp4", 512 | "pix_fmt": "yuv420p", 513 | "crf": 19, 514 | "save_metadata": false, 515 | "trim_to_audio": false, 516 | "pingpong": false, 517 | "save_output": false, 518 | "videopreview": { 519 | "hidden": false, 520 | "paused": false, 521 | "params": { 522 | "filename": "AnimateDiff_00011.mp4", 523 | "subfolder": "", 524 | "type": "temp", 525 | "format": "video/h264-mp4", 526 | "frame_rate": 24, 527 | "workflow": "AnimateDiff_00011.png", 528 | "fullpath": "/workspace/ComfyUI/temp/AnimateDiff_00011.mp4" 529 | }, 530 | "muted": false 531 | } 532 | } 533 | }, 534 | { 535 | "id": 45, 536 | "type": "EmptyHunyuanLatentVideo", 537 | "pos": [ 538 | 475.540771484375, 539 | 432.673583984375 540 | ], 541 | "size": [ 542 | 315, 543 | 130 544 | ], 545 | "flags": {}, 546 | "order": 4, 547 | "mode": 0, 548 | "inputs": [], 549 | "outputs": [ 550 | { 551 | "name": "LATENT", 552 | "type": "LATENT", 553 | "links": [ 554 | 180, 555 | 220 556 | ], 557 | "slot_index": 0 558 | } 559 | ], 560 | "properties": { 561 | "Node name for S&R": "EmptyHunyuanLatentVideo" 562 | }, 563 | "widgets_values": [ 564 | 848, 565 | 480, 566 | 49, 567 | 1 568 | ] 569 | }, 570 | { 571 | "id": 95, 572 | "type": "CLIPTextEncode", 573 | "pos": [ 574 | 580, 575 | -1680 576 | ], 577 | "size": [ 578 | 422.84503173828125, 579 | 164.31304931640625 580 | ], 581 | "flags": {}, 582 | "order": 16, 583 | "mode": 0, 584 | "inputs": [ 585 | { 586 | "name": "clip", 587 | "type": "CLIP", 588 | "link": 243 589 | } 590 | ], 591 | "outputs": [ 592 | { 593 | "name": "CONDITIONING", 594 | "type": "CONDITIONING", 595 | "links": [ 596 | 238 597 | ], 598 | "slot_index": 0 599 | } 600 | ], 601 | "title": "CLIP Text Encode (Positive Prompt)", 602 | "properties": { 603 | "Node name for S&R": "CLIPTextEncode" 604 | }, 605 | "widgets_values": [ 606 | "an hd video of a city street scene with flying ships in the sky" 607 | ], 608 | "color": "#232", 609 | "bgcolor": "#353" 610 | }, 611 | { 612 | "id": 44, 613 | "type": "CLIPTextEncode", 614 | "pos": [ 615 | 420, 616 | 200 617 | ], 618 | "size": [ 619 | 422.84503173828125, 620 | 164.31304931640625 621 | ], 622 | "flags": {}, 623 | "order": 13, 624 | "mode": 0, 625 | "inputs": [ 626 | { 627 | "name": "clip", 628 | "type": "CLIP", 629 | "link": 205 630 | } 631 | ], 632 | "outputs": [ 633 | { 634 | "name": "CONDITIONING", 635 | "type": "CONDITIONING", 636 | "links": [ 637 | 175 638 | ], 639 | "slot_index": 0 640 | } 641 | ], 642 | "title": "CLIP Text Encode (Positive Prompt)", 643 | "properties": { 644 | "Node name for S&R": "CLIPTextEncode" 645 | }, 646 | "widgets_values": [ 647 | "a biblical scene" 648 | ], 649 | "color": "#232", 650 | "bgcolor": "#353" 651 | }, 652 | { 653 | "id": 80, 654 | "type": "CLIPTextEncode", 655 | "pos": [ 656 | 544.9306030273438, 657 | -578.5970458984375 658 | ], 659 | "size": [ 660 | 422.84503173828125, 661 | 164.31304931640625 662 | ], 663 | "flags": {}, 664 | "order": 14, 665 | "mode": 0, 666 | "inputs": [ 667 | { 668 | "name": "clip", 669 | "type": "CLIP", 670 | "link": 225 671 | } 672 | ], 673 | "outputs": [ 674 | { 675 | "name": "CONDITIONING", 676 | "type": "CONDITIONING", 677 | "links": [ 678 | 216 679 | ], 680 | "slot_index": 0 681 | } 682 | ], 683 | "title": "CLIP Text Encode (Positive Prompt)", 684 | "properties": { 685 | "Node name for S&R": "CLIPTextEncode" 686 | }, 687 | "widgets_values": [ 688 | "An hd tracking video following a mystical figure sitting in a holy seashell, he is reaching his large arm and hand out to the left. The background features an angelic scene with the heavens opening up. The camera zooms in on his hand as it inches closer to the center" 689 | ], 690 | "color": "#232", 691 | "bgcolor": "#353" 692 | }, 693 | { 694 | "id": 85, 695 | "type": "CLIPTextEncode", 696 | "pos": [ 697 | 586.3402099609375, 698 | -1112.4017333984375 699 | ], 700 | "size": [ 701 | 422.84503173828125, 702 | 164.31304931640625 703 | ], 704 | "flags": {}, 705 | "order": 15, 706 | "mode": 0, 707 | "inputs": [ 708 | { 709 | "name": "clip", 710 | "type": "CLIP", 711 | "link": 242 712 | } 713 | ], 714 | "outputs": [ 715 | { 716 | "name": "CONDITIONING", 717 | "type": "CONDITIONING", 718 | "links": [ 719 | 228 720 | ], 721 | "slot_index": 0 722 | } 723 | ], 724 | "title": "CLIP Text Encode (Positive Prompt)", 725 | "properties": { 726 | "Node name for S&R": "CLIPTextEncode" 727 | }, 728 | "widgets_values": [ 729 | "An hd tracking shot of a desperate man reaching up and to the right. The background features a hell opening up below trying to pull the man down. The camera zooms in on his hand as it inches closer to the center" 730 | ], 731 | "color": "#232", 732 | "bgcolor": "#353" 733 | }, 734 | { 735 | "id": 25, 736 | "type": "RandomNoise", 737 | "pos": [ 738 | 479, 739 | 618 740 | ], 741 | "size": [ 742 | 315, 743 | 82 744 | ], 745 | "flags": {}, 746 | "order": 5, 747 | "mode": 0, 748 | "inputs": [], 749 | "outputs": [ 750 | { 751 | "name": "NOISE", 752 | "type": "NOISE", 753 | "links": [ 754 | 37 755 | ], 756 | "shape": 3 757 | } 758 | ], 759 | "properties": { 760 | "Node name for S&R": "RandomNoise" 761 | }, 762 | "widgets_values": [ 763 | 4, 764 | "fixed" 765 | ], 766 | "color": "#2a363b", 767 | "bgcolor": "#3f5159" 768 | }, 769 | { 770 | "id": 93, 771 | "type": "CreateShapeMask", 772 | "pos": [ 773 | 630, 774 | -2000 775 | ], 776 | "size": [ 777 | 315, 778 | 270 779 | ], 780 | "flags": {}, 781 | "order": 6, 782 | "mode": 0, 783 | "inputs": [], 784 | "outputs": [ 785 | { 786 | "name": "mask", 787 | "type": "MASK", 788 | "links": [ 789 | 237 790 | ], 791 | "slot_index": 0 792 | }, 793 | { 794 | "name": "mask_inverted", 795 | "type": "MASK", 796 | "links": [ 797 | 251 798 | ], 799 | "slot_index": 1 800 | } 801 | ], 802 | "properties": { 803 | "Node name for S&R": "CreateShapeMask" 804 | }, 805 | "widgets_values": [ 806 | "square", 807 | 1, 808 | 48, 809 | 48, 810 | 0, 811 | 848, 812 | 480, 813 | 4000, 814 | 4000 815 | ] 816 | }, 817 | { 818 | "id": 83, 819 | "type": "HYApplyRegionalConds", 820 | "pos": [ 821 | 1190, 822 | -480 823 | ], 824 | "size": [ 825 | 315, 826 | 210 827 | ], 828 | "flags": {}, 829 | "order": 28, 830 | "mode": 0, 831 | "inputs": [ 832 | { 833 | "name": "model", 834 | "type": "MODEL", 835 | "link": 223 836 | }, 837 | { 838 | "name": "cond", 839 | "type": "CONDITIONING", 840 | "link": 226 841 | }, 842 | { 843 | "name": "region_conds", 844 | "type": "REGION_COND", 845 | "link": 260 846 | }, 847 | { 848 | "name": "latent", 849 | "type": "LATENT", 850 | "link": 220 851 | }, 852 | { 853 | "name": "attn_override", 854 | "type": "ATTN_OVERRIDE", 855 | "link": 274, 856 | "shape": 7 857 | } 858 | ], 859 | "outputs": [ 860 | { 861 | "name": "MODEL", 862 | "type": "MODEL", 863 | "links": [ 864 | 224 865 | ], 866 | "slot_index": 0 867 | } 868 | ], 869 | "properties": { 870 | "Node name for S&R": "HYApplyRegionalConds" 871 | }, 872 | "widgets_values": [ 873 | 0, 874 | 0.5, 875 | 0, 876 | false 877 | ] 878 | }, 879 | { 880 | "id": 101, 881 | "type": "HYAttnOverride", 882 | "pos": [ 883 | 1610.6875, 884 | -444.72222900390625 885 | ], 886 | "size": [ 887 | 400, 888 | 200 889 | ], 890 | "flags": {}, 891 | "order": 7, 892 | "mode": 4, 893 | "inputs": [], 894 | "outputs": [ 895 | { 896 | "name": "ATTN_OVERRIDE", 897 | "type": "ATTN_OVERRIDE", 898 | "links": [ 899 | 274 900 | ], 901 | "slot_index": 0 902 | } 903 | ], 904 | "properties": { 905 | "Node name for S&R": "HYAttnOverride" 906 | }, 907 | "widgets_values": [ 908 | "0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30", 909 | "0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30" 910 | ] 911 | }, 912 | { 913 | "id": 100, 914 | "type": "Note", 915 | "pos": [ 916 | 2053.74267578125, 917 | -439.9007568359375 918 | ], 919 | "size": [ 920 | 322.08740234375, 921 | 305.0302734375 922 | ], 923 | "flags": {}, 924 | "order": 8, 925 | "mode": 0, 926 | "inputs": [], 927 | "outputs": [], 928 | "properties": {}, 929 | "widgets_values": [ 930 | "---- Attention Override (optional) ----\nYou can adjust the strength of the regions by adding attention overrides.\n\nToo strong and it becomes separate videos, but too weak and it has little change.\n\n\n---- main_cond_strength ----\nThe main prompt is very strong and can completely remove regions. Turn this very low or even to 0.\nAt 0 it is turned off completely.\n\nHowever, at 0 you need to make sure the regional prompts cover all areas of the video or you will get strange grid patterns in regions that are not prompted." 931 | ], 932 | "color": "#432", 933 | "bgcolor": "#653" 934 | }, 935 | { 936 | "id": 86, 937 | "type": "CreateShapeMask", 938 | "pos": [ 939 | 636.2345581054688, 940 | -1429.23583984375 941 | ], 942 | "size": [ 943 | 315, 944 | 270 945 | ], 946 | "flags": {}, 947 | "order": 9, 948 | "mode": 0, 949 | "inputs": [], 950 | "outputs": [ 951 | { 952 | "name": "mask", 953 | "type": "MASK", 954 | "links": [ 955 | 233 956 | ], 957 | "slot_index": 0 958 | }, 959 | { 960 | "name": "mask_inverted", 961 | "type": "MASK", 962 | "links": [ 963 | 276 964 | ], 965 | "slot_index": 1 966 | } 967 | ], 968 | "properties": { 969 | "Node name for S&R": "CreateShapeMask" 970 | }, 971 | "widgets_values": [ 972 | "square", 973 | 1, 974 | 848, 975 | 0, 976 | 0, 977 | 848, 978 | 480, 979 | 848, 980 | 4096 981 | ] 982 | }, 983 | { 984 | "id": 81, 985 | "type": "CreateShapeMask", 986 | "pos": [ 987 | 594.824951171875, 988 | -895.43115234375 989 | ], 990 | "size": [ 991 | 315, 992 | 270 993 | ], 994 | "flags": {}, 995 | "order": 10, 996 | "mode": 0, 997 | "inputs": [], 998 | "outputs": [ 999 | { 1000 | "name": "mask", 1001 | "type": "MASK", 1002 | "links": [ 1003 | 232 1004 | ], 1005 | "slot_index": 0 1006 | }, 1007 | { 1008 | "name": "mask_inverted", 1009 | "type": "MASK", 1010 | "links": [ 1011 | 277 1012 | ], 1013 | "slot_index": 1 1014 | } 1015 | ], 1016 | "properties": { 1017 | "Node name for S&R": "CreateShapeMask" 1018 | }, 1019 | "widgets_values": [ 1020 | "square", 1021 | 1, 1022 | 0, 1023 | 0, 1024 | 0, 1025 | 848, 1026 | 480, 1027 | 848, 1028 | 1000 1029 | ] 1030 | }, 1031 | { 1032 | "id": 91, 1033 | "type": "PreviewImage", 1034 | "pos": [ 1035 | 1550.2379150390625, 1036 | -856.0711059570312 1037 | ], 1038 | "size": [ 1039 | 210, 1040 | 246 1041 | ], 1042 | "flags": {}, 1043 | "order": 26, 1044 | "mode": 0, 1045 | "inputs": [ 1046 | { 1047 | "name": "images", 1048 | "type": "IMAGE", 1049 | "link": 234 1050 | } 1051 | ], 1052 | "outputs": [], 1053 | "properties": { 1054 | "Node name for S&R": "PreviewImage" 1055 | }, 1056 | "widgets_values": [] 1057 | }, 1058 | { 1059 | "id": 90, 1060 | "type": "PreviewImage", 1061 | "pos": [ 1062 | 1696.442626953125, 1063 | -1371.86572265625 1064 | ], 1065 | "size": [ 1066 | 210, 1067 | 246 1068 | ], 1069 | "flags": {}, 1070 | "order": 25, 1071 | "mode": 0, 1072 | "inputs": [ 1073 | { 1074 | "name": "images", 1075 | "type": "IMAGE", 1076 | "link": 235 1077 | } 1078 | ], 1079 | "outputs": [], 1080 | "properties": { 1081 | "Node name for S&R": "PreviewImage" 1082 | }, 1083 | "widgets_values": [] 1084 | }, 1085 | { 1086 | "id": 87, 1087 | "type": "HYCreateRegionalCond", 1088 | "pos": [ 1089 | 1169.326416015625, 1090 | -1032.211181640625 1091 | ], 1092 | "size": [ 1093 | 304.79998779296875, 1094 | 122 1095 | ], 1096 | "flags": {}, 1097 | "order": 22, 1098 | "mode": 0, 1099 | "inputs": [ 1100 | { 1101 | "name": "cond", 1102 | "type": "CONDITIONING", 1103 | "link": 228 1104 | }, 1105 | { 1106 | "name": "mask", 1107 | "type": "MASK", 1108 | "link": 276 1109 | }, 1110 | { 1111 | "name": "prev_regions", 1112 | "type": "REGION_COND", 1113 | "link": null, 1114 | "shape": 7 1115 | } 1116 | ], 1117 | "outputs": [ 1118 | { 1119 | "name": "REGION_COND", 1120 | "type": "REGION_COND", 1121 | "links": [ 1122 | 236 1123 | ], 1124 | "slot_index": 0 1125 | } 1126 | ], 1127 | "properties": { 1128 | "Node name for S&R": "HYCreateRegionalCond" 1129 | }, 1130 | "widgets_values": [ 1131 | 1, 1132 | "first_only" 1133 | ] 1134 | }, 1135 | { 1136 | "id": 89, 1137 | "type": "MaskToImage", 1138 | "pos": [ 1139 | 1200.27490234375, 1140 | -848.5791625976562 1141 | ], 1142 | "size": [ 1143 | 264.5999755859375, 1144 | 26 1145 | ], 1146 | "flags": {}, 1147 | "order": 19, 1148 | "mode": 0, 1149 | "inputs": [ 1150 | { 1151 | "name": "mask", 1152 | "type": "MASK", 1153 | "link": 232 1154 | } 1155 | ], 1156 | "outputs": [ 1157 | { 1158 | "name": "IMAGE", 1159 | "type": "IMAGE", 1160 | "links": [ 1161 | 234 1162 | ], 1163 | "slot_index": 0 1164 | } 1165 | ], 1166 | "properties": { 1167 | "Node name for S&R": "MaskToImage" 1168 | }, 1169 | "widgets_values": [] 1170 | }, 1171 | { 1172 | "id": 88, 1173 | "type": "MaskToImage", 1174 | "pos": [ 1175 | 1173.0250244140625, 1176 | -1133.0557861328125 1177 | ], 1178 | "size": [ 1179 | 264.5999755859375, 1180 | 26 1181 | ], 1182 | "flags": {}, 1183 | "order": 18, 1184 | "mode": 0, 1185 | "inputs": [ 1186 | { 1187 | "name": "mask", 1188 | "type": "MASK", 1189 | "link": 233 1190 | } 1191 | ], 1192 | "outputs": [ 1193 | { 1194 | "name": "IMAGE", 1195 | "type": "IMAGE", 1196 | "links": [ 1197 | 235 1198 | ], 1199 | "slot_index": 0 1200 | } 1201 | ], 1202 | "properties": { 1203 | "Node name for S&R": "MaskToImage" 1204 | }, 1205 | "widgets_values": [] 1206 | }, 1207 | { 1208 | "id": 94, 1209 | "type": "HYCreateRegionalCond", 1210 | "pos": [ 1211 | 1130.186767578125, 1212 | -1657.6357421875 1213 | ], 1214 | "size": [ 1215 | 304.79998779296875, 1216 | 122 1217 | ], 1218 | "flags": {}, 1219 | "order": 23, 1220 | "mode": 0, 1221 | "inputs": [ 1222 | { 1223 | "name": "cond", 1224 | "type": "CONDITIONING", 1225 | "link": 238 1226 | }, 1227 | { 1228 | "name": "mask", 1229 | "type": "MASK", 1230 | "link": 251 1231 | }, 1232 | { 1233 | "name": "prev_regions", 1234 | "type": "REGION_COND", 1235 | "link": null, 1236 | "shape": 7 1237 | } 1238 | ], 1239 | "outputs": [ 1240 | { 1241 | "name": "REGION_COND", 1242 | "type": "REGION_COND", 1243 | "links": [], 1244 | "slot_index": 0 1245 | } 1246 | ], 1247 | "properties": { 1248 | "Node name for S&R": "HYCreateRegionalCond" 1249 | }, 1250 | "widgets_values": [ 1251 | 0.4, 1252 | "first_only" 1253 | ] 1254 | }, 1255 | { 1256 | "id": 92, 1257 | "type": "MaskToImage", 1258 | "pos": [ 1259 | 1159.6939697265625, 1260 | -1745.0765380859375 1261 | ], 1262 | "size": [ 1263 | 264.5999755859375, 1264 | 26 1265 | ], 1266 | "flags": {}, 1267 | "order": 17, 1268 | "mode": 0, 1269 | "inputs": [ 1270 | { 1271 | "name": "mask", 1272 | "type": "MASK", 1273 | "link": 237 1274 | } 1275 | ], 1276 | "outputs": [ 1277 | { 1278 | "name": "IMAGE", 1279 | "type": "IMAGE", 1280 | "links": [ 1281 | 241 1282 | ], 1283 | "slot_index": 0 1284 | } 1285 | ], 1286 | "properties": { 1287 | "Node name for S&R": "MaskToImage" 1288 | }, 1289 | "widgets_values": [] 1290 | }, 1291 | { 1292 | "id": 96, 1293 | "type": "PreviewImage", 1294 | "pos": [ 1295 | 1502.1038818359375, 1296 | -1816.5303955078125 1297 | ], 1298 | "size": [ 1299 | 210, 1300 | 246 1301 | ], 1302 | "flags": {}, 1303 | "order": 24, 1304 | "mode": 0, 1305 | "inputs": [ 1306 | { 1307 | "name": "images", 1308 | "type": "IMAGE", 1309 | "link": 241 1310 | } 1311 | ], 1312 | "outputs": [], 1313 | "properties": { 1314 | "Node name for S&R": "PreviewImage" 1315 | }, 1316 | "widgets_values": [] 1317 | }, 1318 | { 1319 | "id": 78, 1320 | "type": "HYCreateRegionalCond", 1321 | "pos": [ 1322 | 1192.2176513671875, 1323 | -760.3463745117188 1324 | ], 1325 | "size": [ 1326 | 304.79998779296875, 1327 | 122 1328 | ], 1329 | "flags": {}, 1330 | "order": 27, 1331 | "mode": 0, 1332 | "inputs": [ 1333 | { 1334 | "name": "cond", 1335 | "type": "CONDITIONING", 1336 | "link": 216 1337 | }, 1338 | { 1339 | "name": "mask", 1340 | "type": "MASK", 1341 | "link": 277 1342 | }, 1343 | { 1344 | "name": "prev_regions", 1345 | "type": "REGION_COND", 1346 | "link": 236, 1347 | "shape": 7 1348 | } 1349 | ], 1350 | "outputs": [ 1351 | { 1352 | "name": "REGION_COND", 1353 | "type": "REGION_COND", 1354 | "links": [ 1355 | 260 1356 | ], 1357 | "slot_index": 0 1358 | } 1359 | ], 1360 | "properties": { 1361 | "Node name for S&R": "HYCreateRegionalCond" 1362 | }, 1363 | "widgets_values": [ 1364 | 1, 1365 | "first_only" 1366 | ] 1367 | } 1368 | ], 1369 | "links": [ 1370 | [ 1371 | 19, 1372 | 16, 1373 | 0, 1374 | 13, 1375 | 2, 1376 | "SAMPLER" 1377 | ], 1378 | [ 1379 | 20, 1380 | 17, 1381 | 0, 1382 | 13, 1383 | 3, 1384 | "SIGMAS" 1385 | ], 1386 | [ 1387 | 30, 1388 | 22, 1389 | 0, 1390 | 13, 1391 | 1, 1392 | "GUIDER" 1393 | ], 1394 | [ 1395 | 37, 1396 | 25, 1397 | 0, 1398 | 13, 1399 | 0, 1400 | "NOISE" 1401 | ], 1402 | [ 1403 | 129, 1404 | 26, 1405 | 0, 1406 | 22, 1407 | 1, 1408 | "CONDITIONING" 1409 | ], 1410 | [ 1411 | 175, 1412 | 44, 1413 | 0, 1414 | 26, 1415 | 0, 1416 | "CONDITIONING" 1417 | ], 1418 | [ 1419 | 180, 1420 | 45, 1421 | 0, 1422 | 13, 1423 | 4, 1424 | "LATENT" 1425 | ], 1426 | [ 1427 | 190, 1428 | 12, 1429 | 0, 1430 | 17, 1431 | 0, 1432 | "MODEL" 1433 | ], 1434 | [ 1435 | 205, 1436 | 11, 1437 | 0, 1438 | 44, 1439 | 0, 1440 | "CLIP" 1441 | ], 1442 | [ 1443 | 210, 1444 | 13, 1445 | 0, 1446 | 73, 1447 | 0, 1448 | "LATENT" 1449 | ], 1450 | [ 1451 | 211, 1452 | 10, 1453 | 0, 1454 | 73, 1455 | 1, 1456 | "VAE" 1457 | ], 1458 | [ 1459 | 216, 1460 | 80, 1461 | 0, 1462 | 78, 1463 | 0, 1464 | "CONDITIONING" 1465 | ], 1466 | [ 1467 | 220, 1468 | 45, 1469 | 0, 1470 | 83, 1471 | 3, 1472 | "LATENT" 1473 | ], 1474 | [ 1475 | 221, 1476 | 12, 1477 | 0, 1478 | 82, 1479 | 0, 1480 | "MODEL" 1481 | ], 1482 | [ 1483 | 222, 1484 | 82, 1485 | 0, 1486 | 67, 1487 | 0, 1488 | "MODEL" 1489 | ], 1490 | [ 1491 | 223, 1492 | 67, 1493 | 0, 1494 | 83, 1495 | 0, 1496 | "MODEL" 1497 | ], 1498 | [ 1499 | 224, 1500 | 83, 1501 | 0, 1502 | 22, 1503 | 0, 1504 | "MODEL" 1505 | ], 1506 | [ 1507 | 225, 1508 | 11, 1509 | 0, 1510 | 80, 1511 | 0, 1512 | "CLIP" 1513 | ], 1514 | [ 1515 | 226, 1516 | 26, 1517 | 0, 1518 | 83, 1519 | 1, 1520 | "CONDITIONING" 1521 | ], 1522 | [ 1523 | 227, 1524 | 73, 1525 | 0, 1526 | 84, 1527 | 0, 1528 | "IMAGE" 1529 | ], 1530 | [ 1531 | 228, 1532 | 85, 1533 | 0, 1534 | 87, 1535 | 0, 1536 | "CONDITIONING" 1537 | ], 1538 | [ 1539 | 232, 1540 | 81, 1541 | 0, 1542 | 89, 1543 | 0, 1544 | "MASK" 1545 | ], 1546 | [ 1547 | 233, 1548 | 86, 1549 | 0, 1550 | 88, 1551 | 0, 1552 | "MASK" 1553 | ], 1554 | [ 1555 | 234, 1556 | 89, 1557 | 0, 1558 | 91, 1559 | 0, 1560 | "IMAGE" 1561 | ], 1562 | [ 1563 | 235, 1564 | 88, 1565 | 0, 1566 | 90, 1567 | 0, 1568 | "IMAGE" 1569 | ], 1570 | [ 1571 | 236, 1572 | 87, 1573 | 0, 1574 | 78, 1575 | 2, 1576 | "REGION_COND" 1577 | ], 1578 | [ 1579 | 237, 1580 | 93, 1581 | 0, 1582 | 92, 1583 | 0, 1584 | "MASK" 1585 | ], 1586 | [ 1587 | 238, 1588 | 95, 1589 | 0, 1590 | 94, 1591 | 0, 1592 | "CONDITIONING" 1593 | ], 1594 | [ 1595 | 241, 1596 | 92, 1597 | 0, 1598 | 96, 1599 | 0, 1600 | "IMAGE" 1601 | ], 1602 | [ 1603 | 242, 1604 | 11, 1605 | 0, 1606 | 85, 1607 | 0, 1608 | "CLIP" 1609 | ], 1610 | [ 1611 | 243, 1612 | 11, 1613 | 0, 1614 | 95, 1615 | 0, 1616 | "CLIP" 1617 | ], 1618 | [ 1619 | 251, 1620 | 93, 1621 | 1, 1622 | 94, 1623 | 1, 1624 | "MASK" 1625 | ], 1626 | [ 1627 | 260, 1628 | 78, 1629 | 0, 1630 | 83, 1631 | 2, 1632 | "REGION_COND" 1633 | ], 1634 | [ 1635 | 274, 1636 | 101, 1637 | 0, 1638 | 83, 1639 | 4, 1640 | "ATTN_OVERRIDE" 1641 | ], 1642 | [ 1643 | 276, 1644 | 86, 1645 | 1, 1646 | 87, 1647 | 1, 1648 | "MASK" 1649 | ], 1650 | [ 1651 | 277, 1652 | 81, 1653 | 1, 1654 | 78, 1655 | 1, 1656 | "MASK" 1657 | ] 1658 | ], 1659 | "groups": [], 1660 | "config": {}, 1661 | "extra": { 1662 | "ds": { 1663 | "scale": 0.24304420967403448, 1664 | "offset": [ 1665 | 1040.0007253377598, 1666 | 1536.193742700761 1667 | ] 1668 | }, 1669 | "groupNodes": {} 1670 | }, 1671 | "version": 0.4 1672 | } --------------------------------------------------------------------------------