├── dragon.png ├── terrarium.png ├── fluxSchenll.ico ├── fluxSchenll.png ├── Flux1-Schnell2.gif ├── interior-1step.png ├── red-whiteTREE.png ├── requirements.txt ├── interior-2steps.png ├── interior-3steps.png ├── interior-4steps.png ├── GenerateImages.py └── README.md /dragon.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/dragon.png -------------------------------------------------------------------------------- /terrarium.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/terrarium.png -------------------------------------------------------------------------------- /fluxSchenll.ico: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/fluxSchenll.ico -------------------------------------------------------------------------------- /fluxSchenll.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/fluxSchenll.png -------------------------------------------------------------------------------- /Flux1-Schnell2.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/Flux1-Schnell2.gif -------------------------------------------------------------------------------- /interior-1step.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/interior-1step.png -------------------------------------------------------------------------------- /red-whiteTREE.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/red-whiteTREE.png -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/requirements.txt -------------------------------------------------------------------------------- /interior-2steps.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/interior-2steps.png -------------------------------------------------------------------------------- /interior-3steps.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/interior-3steps.png -------------------------------------------------------------------------------- /interior-4steps.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/fabiomatricardi/FLUX1-Schenll_TAI-freeAPI/main/interior-4steps.png -------------------------------------------------------------------------------- /GenerateImages.py: -------------------------------------------------------------------------------- 1 | from together import Together 2 | import requests 3 | from PIL import Image 4 | from io import BytesIO 5 | APIK = input('Your Tegether API Key: ') 6 | prompt = input('/imagine: ') 7 | client = Together(api_key=APIK) 8 | try: 9 | response = client.images.generate(prompt=prompt, 10 | model="black-forest-labs/FLUX.1-schnell-Free", 11 | steps=4, 12 | width=1024, 13 | height=768, 14 | n=1) 15 | imurl = response.data[0].url 16 | my_res = requests.get(imurl) 17 | my_img = Image.open(BytesIO(my_res.content)) 18 | my_img.show() 19 | savename = input('Give me a name (end with .png please): ') 20 | my_img.save(savename) 21 | except: 22 | print('Check you API Key is a valid one at https://api.together.ai/settings/api-keys !') 23 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # FLUX1-Schenll_TAI-freeAPI 2 | How to generate Images with Flux.1-Schnell and free API from Tegether.AI 3 | 4 | 5 | 6 | Together AI provide free access with API to Top Notch models. 7 | 8 | ## Flux.1-Schenll is awesome at very small Steps (even only 1 or 4) 9 | 10 | Flux.1-Schenll can be used for free (siwth some limitations) 11 | - remotely with an API Key 12 | - on the web with https://api.together.ai/ 13 | 14 | Free quota: 15 | - 60 Requests Per Minute 16 | - Total 60.000 requests per Months 17 | 18 | Register at https://api.together.ai/ 19 | 20 | 21 | ### What is Flux.1-Schnell 22 | FLUX.1-Schnell is a 12 billion parameter rectified flow transformer designed to generate high-quality images from text descriptions. It is trained using latent adversarial diffusion distillation, which allows it to produce images in just 1 to 4 steps. This model is optimized for speed and efficiency, making it suitable for both personal and commercial use. 23 | 24 | It is available under an Apache 2.0 license, and its implementation and sampling code can be found in a dedicated GitHub repository. FLUX.1-Schnell can be used to create a variety of images, including photorealistic scenes, artistic illustrations, and concept art, applicable to diverse applications. 25 | 26 | FLUX.1-Schnell can produce images in only 4 steps due to its training methodology and architecture. Here are the key factors that enable this efficiency: 27 | 28 | - Latent Adversarial Diffusion Distillation: This training technique allows the model to generate high-quality images with fewer steps compared to traditional diffusion models. It distills the knowledge from a larger, more complex model into a smaller, more efficient one, enabling faster inference. 29 | - Rectified Flow Transformer Architecture: The model's architecture is designed to optimize the flow of information during the generation process. This includes a hybrid architecture that combines multimodal and parallel diffusion transformer blocks, enhancing both speed and quality. 30 | - Optimized Inference: The model includes optimizations such as compiled fp8 quantization and optimized attention kernels, which further accelerate the image generation process. These optimizations are crucial for achieving high-quality outputs in fewer steps. 31 | 32 | #### Note 33 | > You cannot set more than 4 steps in the generation! 34 | 35 | ## Generation quality by number of steps 36 | 37 | > PROMPT: Depict a cozy, warmly lit bookstore cafe on a rainy evening. The atmosphere should be inviting and nostalgic, with soft yellow lighting from vintage lamps illuminating rows of well-worn books. Show patrons reading in comfortable armchairs, steam rising from their coffee cups. The large front window should reveal a glistening wet street outside, with blurred lights from passing cars. Emphasize the contrast between the warm interior and the cool, rainy exterior. 38 | 39 | 40 | Here the same prompt from 1 to 2 steps (left to right) 41 | 42 | 43 | 44 | Here the same prompt from 3 to 4 steps (left to right) 45 | 46 | 47 | 48 | 49 | 50 | 51 | --- 52 | 53 | Remote calls requires the Together package. Recommended to use Python 3.12 to avoid dependency clashes 54 | ``` 55 | pip install together==1.4.1 56 | ``` 57 | 58 | Or Clone the repo and create a venv and install from requirements 59 | ``` 60 | python312 -m venv venv 61 | venv\Scripts\activate 62 | pip install -r requirements.txt 63 | ``` 64 | 65 | To run the generation: 66 | ``` 67 | Python .\GenerateImages.py 68 | ``` 69 | You will be asked for: 70 | - a valid API key 71 | - a prompt 72 | - a filename, once the miage is genrated 73 | 74 | 75 | ### Some prompt examples 76 | You can find more [here](https://getimg.ai/blog/flux-1-prompt-guide-pro-tips-and-common-mistakes-to-avoid) and [here](https://aimlapi.com/blog/master-the-art-of-ai-top-10-prompts-for-flux-1-by-black-forests-labs) 77 | 78 | 79 | > A hanging glass terrarium featuring a miniature rainforest scene with colorful orchids and tiny waterfalls. Just beyond the glass, a neon sign reads 'Rainforest Retreat.' The rain-soaked glass creates a beautiful distortion, adding a soft glow to the sign's vibrant colors 80 | 81 | 82 | 83 | 84 | --- 85 | 86 | 87 | > Photo realistic scene inspired by LOTR: [A tiny red dragon sleeps curled up in a nest on a medieval wizard's table]. Shot with a macro lens (f/2.8, 50mm) and a Canon EOSR5, the soft focus captures [the cozy morning light filtering through a near by window]. The pastel colors and whimsical steam shapes enhance the serene atmosphere, evoking a DnD RPG setting. The image is rendered in 16K and 8K, highlighting [the intricate details and medieval charm]. 88 | 89 | 90 | 91 | --- 92 | 93 | 94 | > A single tree standing in the middle of the image. The left half of the tree has bright, vibrant red leaves under a bright, sunny blue sky, while the right half has bare branches covered in frost, with a cold, dark, thunderous sky. On the left there's orange, lush grass on the ground; on the right - thick snow. The split is sharp, with the transition happening right down the middle of the tree 95 | 96 | 97 | 98 | 99 | --- 100 | 101 | ## Resources 102 | - Together AI [pypi package page](https://pypi.org/project/together/) 103 | - [Together.AI](https://api.together.ai/) 104 | - [Introduction](https://docs.together.ai/docs/introduction) to TogetherAI and services 105 | - [Together.Cookbooks](https://github.com/togethercomputer/together-cookbook) 106 | - How to create Images - [QuickStart](https://docs.together.ai/docs/images-overview) 107 | 108 | 109 | --- 110 | 111 | Gradio Interface soon 112 | 113 | --- 114 | 115 | 116 | 117 | 118 | 119 | 120 | --------------------------------------------------------------------------------