├── LICENSE.txt ├── Line2Normalmap.py ├── Line2Normalmap_ReadMe.txt ├── Line2Normalmap_gui.py ├── Line2Normalmap_install.ps1 ├── Line2Normalmap_model_DL.cmd ├── Line2Normalmap_modules ├── config_states.py ├── gitpython_hack.py ├── launch_utils_Line2Normalmap.py ├── sd1_clip.py ├── sd2_clip.py ├── sdxl_clip.py ├── shared_cmd_options.py └── ui_extensions.py ├── Line2Normalmap_setup.py ├── README.md ├── utils ├── application.py ├── request_api.py └── tagger.py └── venv.cmd /LICENSE.txt: -------------------------------------------------------------------------------- 1 | GNU AFFERO GENERAL PUBLIC LICENSE 2 | Version 3, 19 November 2007 3 | 4 | Copyright (c) 2023 AUTOMATIC1111 5 | 6 | Copyright (C) 2007 Free Software Foundation, Inc. 7 | Everyone is permitted to copy and distribute verbatim copies 8 | of this license document, but changing it is not allowed. 9 | 10 | Preamble 11 | 12 | The GNU Affero General Public License is a free, copyleft license for 13 | software and other kinds of works, specifically designed to ensure 14 | cooperation with the community in the case of network server software. 15 | 16 | The licenses for most software and other practical works are designed 17 | to take away your freedom to share and change the works. By contrast, 18 | our General Public Licenses are intended to guarantee your freedom to 19 | share and change all versions of a program--to make sure it remains free 20 | software for all its users. 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 | Developers that use our General Public Licenses protect your rights 30 | with two steps: (1) assert copyright on the software, and (2) offer 31 | you this License which gives you legal permission to copy, distribute 32 | and/or modify the software. 33 | 34 | A secondary benefit of defending all users' freedom is that 35 | improvements made in alternate versions of the program, if they 36 | receive widespread use, become available for other developers to 37 | incorporate. Many developers of free software are heartened and 38 | encouraged by the resulting cooperation. However, in the case of 39 | software used on network servers, this result may fail to come about. 40 | The GNU General Public License permits making a modified version and 41 | letting the public access it on a server without ever releasing its 42 | source code to the public. 43 | 44 | The GNU Affero General Public License is designed specifically to 45 | ensure that, in such cases, the modified source code becomes available 46 | to the community. It requires the operator of a network server to 47 | provide the source code of the modified version running there to the 48 | users of that server. Therefore, public use of a modified version, on 49 | a publicly accessible server, gives the public access to the source 50 | code of the modified version. 51 | 52 | An older license, called the Affero General Public License and 53 | published by Affero, was designed to accomplish similar goals. This is 54 | a different license, not a version of the Affero GPL, but Affero has 55 | released a new version of the Affero GPL which permits relicensing under 56 | this license. 57 | 58 | The precise terms and conditions for copying, distribution and 59 | modification follow. 60 | 61 | TERMS AND CONDITIONS 62 | 63 | 0. Definitions. 64 | 65 | "This License" refers to version 3 of the GNU Affero General Public License. 66 | 67 | "Copyright" also means copyright-like laws that apply to other kinds of 68 | works, such as semiconductor masks. 69 | 70 | "The Program" refers to any copyrightable work licensed under this 71 | License. Each licensee is addressed as "you". "Licensees" and 72 | "recipients" may be individuals or organizations. 73 | 74 | To "modify" a work means to copy from or adapt all or part of the work 75 | in a fashion requiring copyright permission, other than the making of an 76 | exact copy. The resulting work is called a "modified version" of the 77 | earlier work or a work "based on" the earlier work. 78 | 79 | A "covered work" means either the unmodified Program or a work based 80 | on the Program. 81 | 82 | To "propagate" a work means to do anything with it that, without 83 | permission, would make you directly or secondarily liable for 84 | infringement under applicable copyright law, except executing it on a 85 | computer or modifying a private copy. Propagation includes copying, 86 | distribution (with or without modification), making available to the 87 | public, and in some countries other activities as well. 88 | 89 | To "convey" a work means any kind of propagation that enables other 90 | parties to make or receive copies. Mere interaction with a user through 91 | a computer network, with no transfer of a copy, is not conveying. 92 | 93 | An interactive user interface displays "Appropriate Legal Notices" 94 | to the extent that it includes a convenient and prominently visible 95 | feature that (1) displays an appropriate copyright notice, and (2) 96 | tells the user that there is no warranty for the work (except to the 97 | extent that warranties are provided), that licensees may convey the 98 | work under this License, and how to view a copy of this License. If 99 | the interface presents a list of user commands or options, such as a 100 | menu, a prominent item in the list meets this criterion. 101 | 102 | 1. Source Code. 103 | 104 | The "source code" for a work means the preferred form of the work 105 | for making modifications to it. "Object code" means any non-source 106 | form of a work. 107 | 108 | A "Standard Interface" means an interface that either is an official 109 | standard defined by a recognized standards body, or, in the case of 110 | interfaces specified for a particular programming language, one that 111 | is widely used among developers working in that language. 112 | 113 | The "System Libraries" of an executable work include anything, other 114 | than the work as a whole, that (a) is included in the normal form of 115 | packaging a Major Component, but which is not part of that Major 116 | Component, and (b) serves only to enable use of the work with that 117 | Major Component, or to implement a Standard Interface for which an 118 | implementation is available to the public in source code form. A 119 | "Major Component", in this context, means a major essential component 120 | (kernel, window system, and so on) of the specific operating system 121 | (if any) on which the executable work runs, or a compiler used to 122 | produce the work, or an object code interpreter used to run it. 123 | 124 | The "Corresponding Source" for a work in object code form means all 125 | the source code needed to generate, install, and (for an executable 126 | work) run the object code and to modify the work, including scripts to 127 | control those activities. However, it does not include the work's 128 | System Libraries, or general-purpose tools or generally available free 129 | programs which are used unmodified in performing those activities but 130 | which are not part of the work. For example, Corresponding Source 131 | includes interface definition files associated with source files for 132 | the work, and the source code for shared libraries and dynamically 133 | linked subprograms that the work is specifically designed to require, 134 | such as by intimate data communication or control flow between those 135 | subprograms and other parts of the work. 136 | 137 | The Corresponding Source need not include anything that users 138 | can regenerate automatically from other parts of the Corresponding 139 | Source. 140 | 141 | The Corresponding Source for a work in source code form is that 142 | same work. 143 | 144 | 2. Basic Permissions. 145 | 146 | All rights granted under this License are granted for the term of 147 | copyright on the Program, and are irrevocable provided the stated 148 | conditions are met. This License explicitly affirms your unlimited 149 | permission to run the unmodified Program. The output from running a 150 | covered work is covered by this License only if the output, given its 151 | content, constitutes a covered work. This License acknowledges your 152 | rights of fair use or other equivalent, as provided by copyright law. 153 | 154 | You may make, run and propagate covered works that you do not 155 | convey, without conditions so long as your license otherwise remains 156 | in force. You may convey covered works to others for the sole purpose 157 | of having them make modifications exclusively for you, or provide you 158 | with facilities for running those works, provided that you comply with 159 | the terms of this License in conveying all material for which you do 160 | not control copyright. Those thus making or running the covered works 161 | for you must do so exclusively on your behalf, under your direction 162 | and control, on terms that prohibit them from making any copies of 163 | your copyrighted material outside their relationship with you. 164 | 165 | Conveying under any other circumstances is permitted solely under 166 | the conditions stated below. Sublicensing is not allowed; section 10 167 | makes it unnecessary. 168 | 169 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 170 | 171 | No covered work shall be deemed part of an effective technological 172 | measure under any applicable law fulfilling obligations under article 173 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 174 | similar laws prohibiting or restricting circumvention of such 175 | measures. 176 | 177 | When you convey a covered work, you waive any legal power to forbid 178 | circumvention of technological measures to the extent such circumvention 179 | is effected by exercising rights under this License with respect to 180 | the covered work, and you disclaim any intention to limit operation or 181 | modification of the work as a means of enforcing, against the work's 182 | users, your or third parties' legal rights to forbid circumvention of 183 | technological measures. 184 | 185 | 4. Conveying Verbatim Copies. 186 | 187 | You may convey verbatim copies of the Program's source code as you 188 | receive it, in any medium, provided that you conspicuously and 189 | appropriately publish on each copy an appropriate copyright notice; 190 | keep intact all notices stating that this License and any 191 | non-permissive terms added in accord with section 7 apply to the code; 192 | keep intact all notices of the absence of any warranty; and give all 193 | recipients a copy of this License along with the Program. 194 | 195 | You may charge any price or no price for each copy that you convey, 196 | and you may offer support or warranty protection for a fee. 197 | 198 | 5. Conveying Modified Source Versions. 199 | 200 | You may convey a work based on the Program, or the modifications to 201 | produce it from the Program, in the form of source code under the 202 | terms of section 4, provided that you also meet all of these conditions: 203 | 204 | a) The work must carry prominent notices stating that you modified 205 | it, and giving a relevant date. 206 | 207 | b) The work must carry prominent notices stating that it is 208 | released under this License and any conditions added under section 209 | 7. This requirement modifies the requirement in section 4 to 210 | "keep intact all notices". 211 | 212 | c) You must license the entire work, as a whole, under this 213 | License to anyone who comes into possession of a copy. This 214 | License will therefore apply, along with any applicable section 7 215 | additional terms, to the whole of the work, and all its parts, 216 | regardless of how they are packaged. This License gives no 217 | permission to license the work in any other way, but it does not 218 | invalidate such permission if you have separately received it. 219 | 220 | d) If the work has interactive user interfaces, each must display 221 | Appropriate Legal Notices; however, if the Program has interactive 222 | interfaces that do not display Appropriate Legal Notices, your 223 | work need not make them do so. 224 | 225 | A compilation of a covered work with other separate and independent 226 | works, which are not by their nature extensions of the covered work, 227 | and which are not combined with it such as to form a larger program, 228 | in or on a volume of a storage or distribution medium, is called an 229 | "aggregate" if the compilation and its resulting copyright are not 230 | used to limit the access or legal rights of the compilation's users 231 | beyond what the individual works permit. Inclusion of a covered work 232 | in an aggregate does not cause this License to apply to the other 233 | parts of the aggregate. 234 | 235 | 6. Conveying Non-Source Forms. 236 | 237 | You may convey a covered work in object code form under the terms 238 | of sections 4 and 5, provided that you also convey the 239 | machine-readable Corresponding Source under the terms of this License, 240 | in one of these ways: 241 | 242 | a) Convey the object code in, or embodied in, a physical product 243 | (including a physical distribution medium), accompanied by the 244 | Corresponding Source fixed on a durable physical medium 245 | customarily used for software interchange. 246 | 247 | b) Convey the object code in, or embodied in, a physical product 248 | (including a physical distribution medium), accompanied by a 249 | written offer, valid for at least three years and valid for as 250 | long as you offer spare parts or customer support for that product 251 | model, to give anyone who possesses the object code either (1) a 252 | copy of the Corresponding Source for all the software in the 253 | product that is covered by this License, on a durable physical 254 | medium customarily used for software interchange, for a price no 255 | more than your reasonable cost of physically performing this 256 | conveying of source, or (2) access to copy the 257 | Corresponding Source from a network server at no charge. 258 | 259 | c) Convey individual copies of the object code with a copy of the 260 | written offer to provide the Corresponding Source. This 261 | alternative is allowed only occasionally and noncommercially, and 262 | only if you received the object code with such an offer, in accord 263 | with subsection 6b. 264 | 265 | d) Convey the object code by offering access from a designated 266 | place (gratis or for a charge), and offer equivalent access to the 267 | Corresponding Source in the same way through the same place at no 268 | further charge. You need not require recipients to copy the 269 | Corresponding Source along with the object code. If the place to 270 | copy the object code is a network server, the Corresponding Source 271 | may be on a different server (operated by you or a third party) 272 | that supports equivalent copying facilities, provided you maintain 273 | clear directions next to the object code saying where to find the 274 | Corresponding Source. Regardless of what server hosts the 275 | Corresponding Source, you remain obligated to ensure that it is 276 | available for as long as needed to satisfy these requirements. 277 | 278 | e) Convey the object code using peer-to-peer transmission, provided 279 | you inform other peers where the object code and Corresponding 280 | Source of the work are being offered to the general public at no 281 | charge under subsection 6d. 282 | 283 | A separable portion of the object code, whose source code is excluded 284 | from the Corresponding Source as a System Library, need not be 285 | included in conveying the object code work. 286 | 287 | A "User Product" is either (1) a "consumer product", which means any 288 | tangible personal property which is normally used for personal, family, 289 | or household purposes, or (2) anything designed or sold for incorporation 290 | into a dwelling. In determining whether a product is a consumer product, 291 | doubtful cases shall be resolved in favor of coverage. For a particular 292 | product received by a particular user, "normally used" refers to a 293 | typical or common use of that class of product, regardless of the status 294 | of the particular user or of the way in which the particular user 295 | actually uses, or expects or is expected to use, the product. A product 296 | is a consumer product regardless of whether the product has substantial 297 | commercial, industrial or non-consumer uses, unless such uses represent 298 | the only significant mode of use of the product. 299 | 300 | "Installation Information" for a User Product means any methods, 301 | procedures, authorization keys, or other information required to install 302 | and execute modified versions of a covered work in that User Product from 303 | a modified version of its Corresponding Source. The information must 304 | suffice to ensure that the continued functioning of the modified object 305 | code is in no case prevented or interfered with solely because 306 | modification has been made. 307 | 308 | If you convey an object code work under this section in, or with, or 309 | specifically for use in, a User Product, and the conveying occurs as 310 | part of a transaction in which the right of possession and use of the 311 | User Product is transferred to the recipient in perpetuity or for a 312 | fixed term (regardless of how the transaction is characterized), the 313 | Corresponding Source conveyed under this section must be accompanied 314 | by the Installation Information. But this requirement does not apply 315 | if neither you nor any third party retains the ability to install 316 | modified object code on the User Product (for example, the work has 317 | been installed in ROM). 318 | 319 | The requirement to provide Installation Information does not include a 320 | requirement to continue to provide support service, warranty, or updates 321 | for a work that has been modified or installed by the recipient, or for 322 | the User Product in which it has been modified or installed. Access to a 323 | network may be denied when the modification itself materially and 324 | adversely affects the operation of the network or violates the rules and 325 | protocols for communication across the network. 326 | 327 | Corresponding Source conveyed, and Installation Information provided, 328 | in accord with this section must be in a format that is publicly 329 | documented (and with an implementation available to the public in 330 | source code form), and must require no special password or key for 331 | unpacking, reading or copying. 332 | 333 | 7. Additional Terms. 334 | 335 | "Additional permissions" are terms that supplement the terms of this 336 | License by making exceptions from one or more of its conditions. 337 | Additional permissions that are applicable to the entire Program shall 338 | be treated as though they were included in this License, to the extent 339 | that they are valid under applicable law. If additional permissions 340 | apply only to part of the Program, that part may be used separately 341 | under those permissions, but the entire Program remains governed by 342 | this License without regard to the additional permissions. 343 | 344 | When you convey a copy of a covered work, you may at your option 345 | remove any additional permissions from that copy, or from any part of 346 | it. (Additional permissions may be written to require their own 347 | removal in certain cases when you modify the work.) You may place 348 | additional permissions on material, added by you to a covered work, 349 | for which you have or can give appropriate copyright permission. 350 | 351 | Notwithstanding any other provision of this License, for material you 352 | add to a covered work, you may (if authorized by the copyright holders of 353 | that material) supplement the terms of this License with terms: 354 | 355 | a) Disclaiming warranty or limiting liability differently from the 356 | terms of sections 15 and 16 of this License; or 357 | 358 | b) Requiring preservation of specified reasonable legal notices or 359 | author attributions in that material or in the Appropriate Legal 360 | Notices displayed by works containing it; or 361 | 362 | c) Prohibiting misrepresentation of the origin of that material, or 363 | requiring that modified versions of such material be marked in 364 | reasonable ways as different from the original version; or 365 | 366 | d) Limiting the use for publicity purposes of names of licensors or 367 | authors of the material; or 368 | 369 | e) Declining to grant rights under trademark law for use of some 370 | trade names, trademarks, or service marks; or 371 | 372 | f) Requiring indemnification of licensors and authors of that 373 | material by anyone who conveys the material (or modified versions of 374 | it) with contractual assumptions of liability to the recipient, for 375 | any liability that these contractual assumptions directly impose on 376 | those licensors and authors. 377 | 378 | All other non-permissive additional terms are considered "further 379 | restrictions" within the meaning of section 10. If the Program as you 380 | received it, or any part of it, contains a notice stating that it is 381 | governed by this License along with a term that is a further 382 | restriction, you may remove that term. If a license document contains 383 | a further restriction but permits relicensing or conveying under this 384 | License, you may add to a covered work material governed by the terms 385 | of that license document, provided that the further restriction does 386 | not survive such relicensing or conveying. 387 | 388 | If you add terms to a covered work in accord with this section, you 389 | must place, in the relevant source files, a statement of the 390 | additional terms that apply to those files, or a notice indicating 391 | where to find the applicable terms. 392 | 393 | Additional terms, permissive or non-permissive, may be stated in the 394 | form of a separately written license, or stated as exceptions; 395 | the above requirements apply either way. 396 | 397 | 8. Termination. 398 | 399 | You may not propagate or modify a covered work except as expressly 400 | provided under this License. Any attempt otherwise to propagate or 401 | modify it is void, and will automatically terminate your rights under 402 | this License (including any patent licenses granted under the third 403 | paragraph of section 11). 404 | 405 | However, if you cease all violation of this License, then your 406 | license from a particular copyright holder is reinstated (a) 407 | provisionally, unless and until the copyright holder explicitly and 408 | finally terminates your license, and (b) permanently, if the copyright 409 | holder fails to notify you of the violation by some reasonable means 410 | prior to 60 days after the cessation. 411 | 412 | Moreover, your license from a particular copyright holder is 413 | reinstated permanently if the copyright holder notifies you of the 414 | violation by some reasonable means, this is the first time you have 415 | received notice of violation of this License (for any work) from that 416 | copyright holder, and you cure the violation prior to 30 days after 417 | your receipt of the notice. 418 | 419 | Termination of your rights under this section does not terminate the 420 | licenses of parties who have received copies or rights from you under 421 | this License. If your rights have been terminated and not permanently 422 | reinstated, you do not qualify to receive new licenses for the same 423 | material under section 10. 424 | 425 | 9. Acceptance Not Required for Having Copies. 426 | 427 | You are not required to accept this License in order to receive or 428 | run a copy of the Program. Ancillary propagation of a covered work 429 | occurring solely as a consequence of using peer-to-peer transmission 430 | to receive a copy likewise does not require acceptance. However, 431 | nothing other than this License grants you permission to propagate or 432 | modify any covered work. These actions infringe copyright if you do 433 | not accept this License. Therefore, by modifying or propagating a 434 | covered work, you indicate your acceptance of this License to do so. 435 | 436 | 10. Automatic Licensing of Downstream Recipients. 437 | 438 | Each time you convey a covered work, the recipient automatically 439 | receives a license from the original licensors, to run, modify and 440 | propagate that work, subject to this License. You are not responsible 441 | for enforcing compliance by third parties with this License. 442 | 443 | An "entity transaction" is a transaction transferring control of an 444 | organization, or substantially all assets of one, or subdividing an 445 | organization, or merging organizations. If propagation of a covered 446 | work results from an entity transaction, each party to that 447 | transaction who receives a copy of the work also receives whatever 448 | licenses to the work the party's predecessor in interest had or could 449 | give under the previous paragraph, plus a right to possession of the 450 | Corresponding Source of the work from the predecessor in interest, if 451 | the predecessor has it or can get it with reasonable efforts. 452 | 453 | You may not impose any further restrictions on the exercise of the 454 | rights granted or affirmed under this License. For example, you may 455 | not impose a license fee, royalty, or other charge for exercise of 456 | rights granted under this License, and you may not initiate litigation 457 | (including a cross-claim or counterclaim in a lawsuit) alleging that 458 | any patent claim is infringed by making, using, selling, offering for 459 | sale, or importing the Program or any portion of it. 460 | 461 | 11. Patents. 462 | 463 | A "contributor" is a copyright holder who authorizes use under this 464 | License of the Program or a work on which the Program is based. The 465 | work thus licensed is called the contributor's "contributor version". 466 | 467 | A contributor's "essential patent claims" are all patent claims 468 | owned or controlled by the contributor, whether already acquired or 469 | hereafter acquired, that would be infringed by some manner, permitted 470 | by this License, of making, using, or selling its contributor version, 471 | but do not include claims that would be infringed only as a 472 | consequence of further modification of the contributor version. For 473 | purposes of this definition, "control" includes the right to grant 474 | patent sublicenses in a manner consistent with the requirements of 475 | this License. 476 | 477 | Each contributor grants you a non-exclusive, worldwide, royalty-free 478 | patent license under the contributor's essential patent claims, to 479 | make, use, sell, offer for sale, import and otherwise run, modify and 480 | propagate the contents of its contributor version. 481 | 482 | In the following three paragraphs, a "patent license" is any express 483 | agreement or commitment, however denominated, not to enforce a patent 484 | (such as an express permission to practice a patent or covenant not to 485 | sue for patent infringement). To "grant" such a patent license to a 486 | party means to make such an agreement or commitment not to enforce a 487 | patent against the party. 488 | 489 | If you convey a covered work, knowingly relying on a patent license, 490 | and the Corresponding Source of the work is not available for anyone 491 | to copy, free of charge and under the terms of this License, through a 492 | publicly available network server or other readily accessible means, 493 | then you must either (1) cause the Corresponding Source to be so 494 | available, or (2) arrange to deprive yourself of the benefit of the 495 | patent license for this particular work, or (3) arrange, in a manner 496 | consistent with the requirements of this License, to extend the patent 497 | license to downstream recipients. "Knowingly relying" means you have 498 | actual knowledge that, but for the patent license, your conveying the 499 | covered work in a country, or your recipient's use of the covered work 500 | in a country, would infringe one or more identifiable patents in that 501 | country that you have reason to believe are valid. 502 | 503 | If, pursuant to or in connection with a single transaction or 504 | arrangement, you convey, or propagate by procuring conveyance of, a 505 | covered work, and grant a patent license to some of the parties 506 | receiving the covered work authorizing them to use, propagate, modify 507 | or convey a specific copy of the covered work, then the patent license 508 | you grant is automatically extended to all recipients of the covered 509 | work and works based on it. 510 | 511 | A patent license is "discriminatory" if it does not include within 512 | the scope of its coverage, prohibits the exercise of, or is 513 | conditioned on the non-exercise of one or more of the rights that are 514 | specifically granted under this License. You may not convey a covered 515 | work if you are a party to an arrangement with a third party that is 516 | in the business of distributing software, under which you make payment 517 | to the third party based on the extent of your activity of conveying 518 | the work, and under which the third party grants, to any of the 519 | parties who would receive the covered work from you, a discriminatory 520 | patent license (a) in connection with copies of the covered work 521 | conveyed by you (or copies made from those copies), or (b) primarily 522 | for and in connection with specific products or compilations that 523 | contain the covered work, unless you entered into that arrangement, 524 | or that patent license was granted, prior to 28 March 2007. 525 | 526 | Nothing in this License shall be construed as excluding or limiting 527 | any implied license or other defenses to infringement that may 528 | otherwise be available to you under applicable patent law. 529 | 530 | 12. No Surrender of Others' Freedom. 531 | 532 | If conditions are imposed on you (whether by court order, agreement or 533 | otherwise) that contradict the conditions of this License, they do not 534 | excuse you from the conditions of this License. If you cannot convey a 535 | covered work so as to satisfy simultaneously your obligations under this 536 | License and any other pertinent obligations, then as a consequence you may 537 | not convey it at all. For example, if you agree to terms that obligate you 538 | to collect a royalty for further conveying from those to whom you convey 539 | the Program, the only way you could satisfy both those terms and this 540 | License would be to refrain entirely from conveying the Program. 541 | 542 | 13. Remote Network Interaction; Use with the GNU General Public License. 543 | 544 | Notwithstanding any other provision of this License, if you modify the 545 | Program, your modified version must prominently offer all users 546 | interacting with it remotely through a computer network (if your version 547 | supports such interaction) an opportunity to receive the Corresponding 548 | Source of your version by providing access to the Corresponding Source 549 | from a network server at no charge, through some standard or customary 550 | means of facilitating copying of software. This Corresponding Source 551 | shall include the Corresponding Source for any work covered by version 3 552 | of the GNU General Public License that is incorporated pursuant to the 553 | following paragraph. 554 | 555 | Notwithstanding any other provision of this License, you have 556 | permission to link or combine any covered work with a work licensed 557 | under version 3 of the GNU General Public License into a single 558 | combined work, and to convey the resulting work. The terms of this 559 | License will continue to apply to the part which is the covered work, 560 | but the work with which it is combined will remain governed by version 561 | 3 of the GNU General Public License. 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 Affero General Public License from time to time. Such new versions 567 | will 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 Affero 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 Affero 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 Affero 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 Affero 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 Affero General Public License for more details. 646 | 647 | You should have received a copy of the GNU Affero 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 your software can interact with users remotely through a computer 653 | network, you should also make sure that it provides a way for users to 654 | get its source. For example, if your program is a web application, its 655 | interface could display a "Source" link that leads users to an archive 656 | of the code. There are many ways you could offer source, and different 657 | solutions will be better for different programs; see section 13 for the 658 | specific requirements. 659 | 660 | You should also get your employer (if you work as a programmer) or school, 661 | if any, to sign a "copyright disclaimer" for the program, if necessary. 662 | For more information on this, and how to apply and follow the GNU AGPL, see 663 | . 664 | -------------------------------------------------------------------------------- /Line2Normalmap.py: -------------------------------------------------------------------------------- 1 | import os 2 | import sys 3 | from modules import launch_utils_Line2Normalmap 4 | 5 | # Default arguments 6 | default_args = ["--nowebui", "--xformers", "--skip-python-version-check", "--skip-torch-cuda-test", "--skip-torch-cuda-test"] 7 | 8 | # Check if custom arguments are provided; if not, append default arguments 9 | if len(sys.argv) == 1: 10 | sys.argv.extend(default_args) 11 | else: 12 | # 独自の引数がある場合、default_argsの中で未指定の引数のみを追加する 13 | # 引数を解析しやすくするため、setを使用 14 | provided_args_set = set(sys.argv) 15 | for arg in default_args: 16 | # "--"で始まるオプションのみを考慮する 17 | if arg.startswith("--"): 18 | option = arg.split("=")[0] if "=" in arg else arg 19 | if option not in provided_args_set: 20 | sys.argv.append(arg) 21 | else: 22 | # "--"で始まらないオプションは直接追加 23 | sys.argv.append(arg) 24 | 25 | args = launch_utils_Line2Normalmap.args 26 | 27 | start = launch_utils_Line2Normalmap.start 28 | 29 | def main(): 30 | start() 31 | 32 | if __name__ == "__main__": 33 | main() -------------------------------------------------------------------------------- /Line2Normalmap_ReadMe.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/tori29umai0123/Line2Normalmap/cbb6a109eddcfe02225e1a2ef1278f6fd9b665e3/Line2Normalmap_ReadMe.txt -------------------------------------------------------------------------------- /Line2Normalmap_gui.py: -------------------------------------------------------------------------------- 1 | from __future__ import annotations 2 | import os 3 | import time 4 | import socket 5 | import atexit 6 | from threading import Thread, Event 7 | import signal 8 | 9 | from modules import timer 10 | from modules import initialize_util 11 | from modules import initialize 12 | from modules_forge.initialization import initialize_forge 13 | from modules_forge import main_thread 14 | from utils import application 15 | 16 | 17 | from uvicorn import Config, Server 18 | import asyncio 19 | 20 | 21 | startup_timer = timer.startup_timer 22 | startup_timer.record("launcher") 23 | 24 | initialize_forge() 25 | initialize.imports() 26 | initialize.check_versions() 27 | initialize.initialize() 28 | 29 | shutdown_event = Event() 30 | 31 | def create_api(app): 32 | from modules.api.api import Api 33 | from modules.call_queue import queue_lock 34 | 35 | api = Api(app, queue_lock) 36 | return api 37 | 38 | def is_port_in_use(port): 39 | with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s: 40 | return s.connect_ex(('localhost', port)) == 0 41 | 42 | def find_available_port(starting_port): 43 | port = starting_port 44 | while is_port_in_use(port): 45 | print(f"Port {port} is in use, trying next one.") 46 | port += 1 47 | return port 48 | 49 | async def api_only_worker(shutdown_event: Event): 50 | from fastapi import FastAPI 51 | import uvicorn 52 | 53 | app = FastAPI() 54 | 55 | app = FastAPI() 56 | initialize_util.setup_middleware(app) 57 | api = create_api(app) 58 | 59 | from modules import script_callbacks 60 | script_callbacks.before_ui_callback() 61 | script_callbacks.app_started_callback(None, app) 62 | 63 | print(f"Startup time: {startup_timer.summary()}.") 64 | 65 | starting_port = 7861 66 | port = find_available_port(starting_port) 67 | 68 | def run_server(): 69 | uvicorn.run(app, host="127.0.0.1", port=port) 70 | 71 | server_thread = Thread(target=run_server) 72 | server_thread.daemon = True 73 | server_thread.start() 74 | 75 | config = uvicorn.Config(app=app, host="127.0.0.1", port={port}, log_level="info") 76 | server = uvicorn.Server(config=config) 77 | 78 | loop = asyncio.get_event_loop() 79 | loop.create_task(server.serve()) 80 | 81 | application.start(f"http://127.0.0.1:{port}") 82 | 83 | shutdown_event.set() 84 | 85 | await os.kill(os.getpid(), signal.SIGTERM) 86 | 87 | def api_only(): 88 | loop = asyncio.get_event_loop() 89 | loop.run_until_complete(api_only_worker(shutdown_event)) 90 | 91 | def on_exit(): 92 | print("Cleaning up...") 93 | shutdown_event.set() 94 | 95 | if __name__ == "__main__": 96 | atexit.register(on_exit) 97 | api_only() 98 | main_thread.loop() -------------------------------------------------------------------------------- /Line2Normalmap_install.ps1: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/tori29umai0123/Line2Normalmap/cbb6a109eddcfe02225e1a2ef1278f6fd9b665e3/Line2Normalmap_install.ps1 -------------------------------------------------------------------------------- /Line2Normalmap_model_DL.cmd: -------------------------------------------------------------------------------- 1 | @echo off 2 | setlocal enabledelayedexpansion 3 | 4 | REM モデルディレクトリの基本パスを実行ディレクトリのmodelsサブディレクトリに設定 5 | set "dpath=%~dp0models" 6 | 7 | REM Taggerモデルダウンロード 8 | set "MODEL_DIR=%dpath%\tagger" 9 | set "MODEL_ID=SmilingWolf/wd-swinv2-tagger-v3" 10 | set "FILES=config.json model.onnx selected_tags.csv sw_jax_cv_config.json" 11 | 12 | if not exist "%MODEL_DIR%" mkdir "%MODEL_DIR%" 13 | 14 | for %%f in (%FILES%) do ( 15 | set "FILE_PATH=%MODEL_DIR%\%%f" 16 | if not exist "!FILE_PATH!" ( 17 | curl -L "https://huggingface.co/%MODEL_ID%/resolve/main/%%f" -o "!FILE_PATH!" 18 | echo Downloaded %%f 19 | ) else ( 20 | echo %%f already exists. 21 | ) 22 | ) 23 | 24 | REM Loraモデルダウンロード 25 | set "MODEL_DIR=%dpath%\Lora" 26 | set "MODEL_ID=tori29umai/SDXL_shadow" 27 | set "FILES=sdxl-testlora-normalmap_04b_dim32.safetensors" 28 | 29 | if not exist "%MODEL_DIR%" mkdir "%MODEL_DIR%" 30 | for %%f in (%FILES%) do ( 31 | set "FILE_PATH=%MODEL_DIR%\%%f" 32 | if not exist "!FILE_PATH!" ( 33 | curl -L "https://huggingface.co/%MODEL_ID%/resolve/main/%%f" -o "!FILE_PATH!" 34 | echo Downloaded %%f 35 | ) else ( 36 | echo %%f already exists. 37 | ) 38 | ) 39 | 40 | REM ControlNetモデルダウンロード 41 | set "MODEL_DIR=%dpath%\ControlNet" 42 | set "MODEL_ID=stabilityai/control-lora" 43 | set "FILES=control-lora-canny-rank256.safetensors" 44 | 45 | if not exist "%MODEL_DIR%" mkdir "%MODEL_DIR%" 46 | for %%f in (%FILES%) do ( 47 | set "FILE_PATH=%MODEL_DIR%\%%f" 48 | if not exist "!FILE_PATH!" ( 49 | curl -L "https://huggingface.co/%MODEL_ID%/resolve/main/control-LoRAs-rank256/%%f" -o "!FILE_PATH!" 50 | echo Downloaded %%f 51 | ) else ( 52 | echo %%f already exists. 53 | ) 54 | ) 55 | 56 | REM Stable-diffusionモデルダウンロード 57 | set "MODEL_DIR=%dpath%\Stable-diffusion" 58 | set "MODEL_ID=cagliostrolab/animagine-xl-3.0" 59 | set "FILES=animagine-xl-3.0.safetensors" 60 | 61 | if not exist "%MODEL_DIR%" mkdir "%MODEL_DIR%" 62 | for %%f in (%FILES%) do ( 63 | set "FILE_PATH=%MODEL_DIR%\%%f" 64 | if not exist "!FILE_PATH!" ( 65 | curl -L "https://huggingface.co/%MODEL_ID%/resolve/main/%%f" -o "!FILE_PATH!" 66 | echo Downloaded %%f 67 | ) else ( 68 | echo %%f already exists. 69 | ) 70 | ) 71 | 72 | endlocal 73 | exit 74 | -------------------------------------------------------------------------------- /Line2Normalmap_modules/config_states.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/tori29umai0123/Line2Normalmap/cbb6a109eddcfe02225e1a2ef1278f6fd9b665e3/Line2Normalmap_modules/config_states.py -------------------------------------------------------------------------------- /Line2Normalmap_modules/gitpython_hack.py: -------------------------------------------------------------------------------- 1 | import io 2 | from typing import Tuple 3 | 4 | import pygit2 5 | 6 | 7 | class Git: 8 | """ 9 | Git wrapper class. 10 | """ 11 | 12 | def __init__(self, repo_path: str): 13 | self.repo_path = repo_path 14 | 15 | def get_object_header(self, ref: str) -> Tuple[str, str, int]: 16 | repo = pygit2.Repository(self.repo_path) 17 | obj = repo.revparse_single(ref) 18 | return obj.hex, obj.type, obj.size 19 | 20 | def stream_object_data(self, ref: str) -> Tuple[str, str, int, io.BytesIO]: 21 | repo = pygit2.Repository(self.repo_path) 22 | obj = repo.revparse_single(ref) 23 | data = obj.data 24 | bio = io.BytesIO(data) 25 | return obj.hex, obj.type, obj.size, bio 26 | 27 | 28 | class Repo: 29 | def __init__(self, repo_path: str): 30 | self.repo_path = repo_path 31 | self.git = Git(repo_path) 32 | -------------------------------------------------------------------------------- /Line2Normalmap_modules/launch_utils_Line2Normalmap.py: -------------------------------------------------------------------------------- 1 | import sys 2 | from modules import cmd_args 3 | 4 | 5 | args, _ = cmd_args.parser.parse_known_args() 6 | 7 | 8 | 9 | def start(): 10 | print(f"Launching {'API server' if '--nowebui' in sys.argv else 'Web UI'} with arguments: {' '.join(sys.argv[1:])}") 11 | import Line2Normalmap_gui 12 | Line2Normalmap_gui.api_only() 13 | 14 | from modules_forge import main_thread 15 | 16 | main_thread.loop() 17 | return 18 | 19 | 20 | def dump_sysinfo(): 21 | from modules import sysinfo 22 | import datetime 23 | 24 | text = sysinfo.get() 25 | filename = f"sysinfo-{datetime.datetime.utcnow().strftime('%Y-%m-%d-%H-%M')}.json" 26 | 27 | with open(filename, "w", encoding="utf8") as file: 28 | file.write(text) 29 | 30 | return filename 31 | -------------------------------------------------------------------------------- /Line2Normalmap_modules/sd1_clip.py: -------------------------------------------------------------------------------- 1 | import sys 2 | # 'frozen' 状態に応じて適切なファイルパスを取得する関数 3 | def get_appropriate_file_path(): 4 | if getattr(sys, 'frozen', False): 5 | return sys.executable + "/Line2Normalmap/" 6 | else: 7 | return __file__ 8 | appropriate_file_path = get_appropriate_file_path() 9 | import sys 10 | # 'frozen'状態に応じて適切なファイルパスを取得する関数 11 | def get_appropriate_file_path(): 12 | if getattr(sys, 'frozen', False): 13 | # ビルドされたアプリケーションの場合、sys.executableのパスを使用 14 | return sys.executable + "/Line2Normalmap/" 15 | else: 16 | # そうでない場合は、従来通りappropriate_file_pathを使用 17 | return appropriate_file_path 18 | 19 | # 適切なファイルパスを取得 20 | appropriate_file_path = get_appropriate_file_path() 21 | 22 | import sys 23 | import os 24 | 25 | # 'frozen'状態に応じて適切なファイルパスを取得する関数 26 | def get_appropriate_file_path(): 27 | if getattr(sys, 'frozen', False): 28 | # ビルドされたアプリケーションの場合、os.path.dirname(sys.executable)のパスを使用 29 | return os.path.dirname(sys.executable) + "/ldm_patched/modules" 30 | else: 31 | # そうでない場合は、従来通りappropriate_file_pathを使用 32 | return os.path.dirname(appropriate_file_path) 33 | 34 | # 適切なファイルパスを取得 35 | appropriate_file_path = get_appropriate_file_path() 36 | 37 | # Implementation of CLIPTextModel transformer 38 | 39 | # using https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py as reference 40 | # written by Forge 41 | 42 | 43 | import os 44 | 45 | from transformers import CLIPTokenizer 46 | import ldm_patched.modules.ops 47 | import torch 48 | import traceback 49 | import zipfile 50 | from . import model_management 51 | import ldm_patched.modules.clip_model 52 | import json 53 | from transformers import CLIPTextModel, CLIPTextConfig, modeling_utils 54 | 55 | 56 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 57 | # This function is only for reference, and not used in the backend or runtime. 58 | def gen_empty_tokens(special_tokens, length): 59 | start_token = special_tokens.get("start", None) 60 | end_token = special_tokens.get("end", None) 61 | pad_token = special_tokens.get("pad") 62 | output = [] 63 | if start_token is not None: 64 | output.append(start_token) 65 | if end_token is not None: 66 | output.append(end_token) 67 | output += [pad_token] * (length - len(output)) 68 | return output 69 | 70 | class ClipTokenWeightEncoder: 71 | 72 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 73 | # This function is only for reference, and not used in the backend or runtime. 74 | def encode_token_weights(self, token_weight_pairs): 75 | to_encode = list() 76 | max_token_len = 0 77 | has_weights = False 78 | for x in token_weight_pairs: 79 | tokens = list(map(lambda a: a[0], x)) 80 | max_token_len = max(len(tokens), max_token_len) 81 | has_weights = has_weights or not all(map(lambda a: a[1] == 1.0, x)) 82 | to_encode.append(tokens) 83 | 84 | sections = len(to_encode) 85 | if has_weights or sections == 0: 86 | to_encode.append(gen_empty_tokens(self.special_tokens, max_token_len)) 87 | 88 | out, pooled = self.encode(to_encode) 89 | if pooled is not None: 90 | first_pooled = pooled[0:1].to(model_management.intermediate_device()) 91 | else: 92 | first_pooled = pooled 93 | 94 | output = [] 95 | for k in range(0, sections): 96 | z = out[k:k+1] 97 | if has_weights: 98 | z_empty = out[-1] 99 | for i in range(len(z)): 100 | for j in range(len(z[i])): 101 | weight = token_weight_pairs[k][j][1] 102 | if weight != 1.0: 103 | z[i][j] = (z[i][j] - z_empty[j]) * weight + z_empty[j] 104 | output.append(z) 105 | 106 | if (len(output) == 0): 107 | return out[-1:].to(model_management.intermediate_device()), first_pooled 108 | return torch.cat(output, dim=-2).to(model_management.intermediate_device()), first_pooled 109 | 110 | class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder): 111 | """Uses the CLIP transformer encoder for text (from huggingface)""" 112 | LAYERS = [ 113 | "last", 114 | "pooled", 115 | "hidden" 116 | ] 117 | def __init__(self, version="openai/clip-vit-large-patch14", device="cpu", max_length=77, 118 | freeze=True, layer="last", layer_idx=None, textmodel_json_config=None, dtype=None, model_class=ldm_patched.modules.clip_model.CLIPTextModel, 119 | special_tokens={"start": 49406, "end": 49407, "pad": 49407}, layer_norm_hidden_state=True): # clip-vit-base-patch32 120 | super().__init__() 121 | assert layer in self.LAYERS 122 | 123 | if textmodel_json_config is None: 124 | textmodel_json_config = os.path.join(appropriate_file_path, "sd1_clip_config.json") 125 | 126 | config = CLIPTextConfig.from_json_file(textmodel_json_config) 127 | self.num_layers = config.num_hidden_layers 128 | 129 | with ldm_patched.modules.ops.use_patched_ops(ldm_patched.modules.ops.manual_cast): 130 | with modeling_utils.no_init_weights(): 131 | self.transformer = CLIPTextModel(config) 132 | 133 | if dtype is not None: 134 | self.transformer.to(dtype) 135 | 136 | self.transformer.text_model.embeddings.to(torch.float32) 137 | 138 | self.max_length = max_length 139 | if freeze: 140 | self.freeze() 141 | self.layer = layer 142 | self.layer_idx = None 143 | self.special_tokens = special_tokens 144 | self.text_projection = torch.nn.Parameter(torch.eye(self.transformer.get_input_embeddings().weight.shape[1])) 145 | self.logit_scale = torch.nn.Parameter(torch.tensor(4.6055)) 146 | self.enable_attention_masks = False 147 | 148 | self.layer_norm_hidden_state = layer_norm_hidden_state 149 | if layer == "hidden": 150 | assert layer_idx is not None 151 | assert abs(layer_idx) < self.num_layers 152 | self.clip_layer(layer_idx) 153 | self.layer_default = (self.layer, self.layer_idx) 154 | 155 | def freeze(self): 156 | self.transformer = self.transformer.eval() 157 | #self.train = disabled_train 158 | for param in self.parameters(): 159 | param.requires_grad = False 160 | 161 | def clip_layer(self, layer_idx): 162 | if abs(layer_idx) > self.num_layers: 163 | self.layer = "last" 164 | else: 165 | self.layer = "hidden" 166 | self.layer_idx = layer_idx 167 | 168 | def reset_clip_layer(self): 169 | self.layer = self.layer_default[0] 170 | self.layer_idx = self.layer_default[1] 171 | 172 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 173 | # This function is only for reference, and not used in the backend or runtime. 174 | def set_up_textual_embeddings(self, tokens, current_embeds): 175 | out_tokens = [] 176 | next_new_token = token_dict_size = current_embeds.weight.shape[0] - 1 177 | embedding_weights = [] 178 | 179 | for x in tokens: 180 | tokens_temp = [] 181 | for y in x: 182 | if isinstance(y, int): 183 | if y == token_dict_size: #EOS token 184 | y = -1 185 | tokens_temp += [y] 186 | else: 187 | if y.shape[0] == current_embeds.weight.shape[1]: 188 | embedding_weights += [y] 189 | tokens_temp += [next_new_token] 190 | next_new_token += 1 191 | else: 192 | print("WARNING: shape mismatch when trying to apply embedding, embedding will be ignored", y.shape[0], current_embeds.weight.shape[1]) 193 | while len(tokens_temp) < len(x): 194 | tokens_temp += [self.special_tokens["pad"]] 195 | out_tokens += [tokens_temp] 196 | 197 | n = token_dict_size 198 | if len(embedding_weights) > 0: 199 | new_embedding = torch.nn.Embedding(next_new_token + 1, current_embeds.weight.shape[1], device=current_embeds.weight.device, dtype=current_embeds.weight.dtype) 200 | new_embedding.weight[:token_dict_size] = current_embeds.weight[:-1] 201 | for x in embedding_weights: 202 | new_embedding.weight[n] = x 203 | n += 1 204 | new_embedding.weight[n] = current_embeds.weight[-1] #EOS embedding 205 | self.transformer.set_input_embeddings(new_embedding) 206 | 207 | processed_tokens = [] 208 | for x in out_tokens: 209 | processed_tokens += [list(map(lambda a: n if a == -1 else a, x))] #The EOS token should always be the largest one 210 | 211 | return processed_tokens 212 | 213 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 214 | # This function is only for reference, and not used in the backend or runtime. 215 | def forward(self, tokens): 216 | backup_embeds = self.transformer.get_input_embeddings() 217 | device = backup_embeds.weight.device 218 | tokens = self.set_up_textual_embeddings(tokens, backup_embeds) 219 | tokens = torch.LongTensor(tokens).to(device) 220 | 221 | attention_mask = None 222 | if self.enable_attention_masks: 223 | attention_mask = torch.zeros_like(tokens) 224 | max_token = self.transformer.get_input_embeddings().weight.shape[0] - 1 225 | for x in range(attention_mask.shape[0]): 226 | for y in range(attention_mask.shape[1]): 227 | attention_mask[x, y] = 1 228 | if tokens[x, y] == max_token: 229 | break 230 | 231 | outputs = self.transformer(input_ids=tokens, attention_mask=attention_mask, 232 | output_hidden_states=self.layer == "hidden") 233 | self.transformer.set_input_embeddings(backup_embeds) 234 | 235 | if self.layer == "last": 236 | z = outputs.last_hidden_state 237 | elif self.layer == "pooled": 238 | z = outputs.pooler_output[:, None, :] 239 | else: 240 | z = outputs.hidden_states[self.layer_idx] 241 | if self.layer_norm_hidden_state: 242 | z = self.transformer.text_model.final_layer_norm(z) 243 | 244 | if hasattr(outputs, "pooler_output"): 245 | pooled_output = outputs.pooler_output.float() 246 | else: 247 | pooled_output = None 248 | 249 | if self.text_projection is not None and pooled_output is not None: 250 | pooled_output = pooled_output.float().to(self.text_projection.device) @ self.text_projection.float() 251 | return z.float(), pooled_output 252 | 253 | def encode(self, tokens): 254 | return self(tokens) 255 | 256 | def load_sd(self, sd): 257 | if "text_projection" in sd: 258 | self.text_projection[:] = sd.pop("text_projection") 259 | if "text_projection.weight" in sd: 260 | self.text_projection[:] = sd.pop("text_projection.weight").transpose(0, 1) 261 | return self.transformer.load_state_dict(sd, strict=False) 262 | 263 | 264 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 265 | # This function is only for reference, and not used in the backend or runtime. 266 | def parse_parentheses(string): 267 | result = [] 268 | current_item = "" 269 | nesting_level = 0 270 | for char in string: 271 | if char == "(": 272 | if nesting_level == 0: 273 | if current_item: 274 | result.append(current_item) 275 | current_item = "(" 276 | else: 277 | current_item = "(" 278 | else: 279 | current_item += char 280 | nesting_level += 1 281 | elif char == ")": 282 | nesting_level -= 1 283 | if nesting_level == 0: 284 | result.append(current_item + ")") 285 | current_item = "" 286 | else: 287 | current_item += char 288 | else: 289 | current_item += char 290 | if current_item: 291 | result.append(current_item) 292 | return result 293 | 294 | 295 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 296 | # This function is only for reference, and not used in the backend or runtime. 297 | def token_weights(string, current_weight): 298 | a = parse_parentheses(string) 299 | out = [] 300 | for x in a: 301 | weight = current_weight 302 | if len(x) >= 2 and x[-1] == ')' and x[0] == '(': 303 | x = x[1:-1] 304 | xx = x.rfind(":") 305 | weight *= 1.1 306 | if xx > 0: 307 | try: 308 | weight = float(x[xx+1:]) 309 | x = x[:xx] 310 | except: 311 | pass 312 | out += token_weights(x, weight) 313 | else: 314 | out += [(x, current_weight)] 315 | return out 316 | 317 | 318 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 319 | # This function is only for reference, and not used in the backend or runtime. 320 | def escape_important(text): 321 | text = text.replace("\\)", "\0\1") 322 | text = text.replace("\\(", "\0\2") 323 | return text 324 | 325 | 326 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 327 | # This function is only for reference, and not used in the backend or runtime. 328 | def unescape_important(text): 329 | text = text.replace("\0\1", ")") 330 | text = text.replace("\0\2", "(") 331 | return text 332 | 333 | 334 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 335 | # This function is only for reference, and not used in the backend or runtime. 336 | def safe_load_embed_zip(embed_path): 337 | with zipfile.ZipFile(embed_path) as myzip: 338 | names = list(filter(lambda a: "data/" in a, myzip.namelist())) 339 | names.reverse() 340 | for n in names: 341 | with myzip.open(n) as myfile: 342 | data = myfile.read() 343 | number = len(data) // 4 344 | length_embed = 1024 #sd2.x 345 | if number < 768: 346 | continue 347 | if number % 768 == 0: 348 | length_embed = 768 #sd1.x 349 | num_embeds = number // length_embed 350 | embed = torch.frombuffer(data, dtype=torch.float) 351 | out = embed.reshape((num_embeds, length_embed)).clone() 352 | del embed 353 | return out 354 | 355 | 356 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 357 | # This function is only for reference, and not used in the backend or runtime. 358 | def expand_directory_list(directories): 359 | dirs = set() 360 | for x in directories: 361 | dirs.add(x) 362 | for root, subdir, file in os.walk(x, followlinks=True): 363 | dirs.add(root) 364 | return list(dirs) 365 | 366 | 367 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 368 | # This function is only for reference, and not used in the backend or runtime. 369 | def load_embed(embedding_name, embedding_directory, embedding_size, embed_key=None): 370 | if isinstance(embedding_directory, str): 371 | embedding_directory = [embedding_directory] 372 | 373 | embedding_directory = expand_directory_list(embedding_directory) 374 | 375 | valid_file = None 376 | for embed_dir in embedding_directory: 377 | embed_path = os.path.abspath(os.path.join(embed_dir, embedding_name)) 378 | embed_dir = os.path.abspath(embed_dir) 379 | try: 380 | if os.path.commonpath((embed_dir, embed_path)) != embed_dir: 381 | continue 382 | except: 383 | continue 384 | if not os.path.isfile(embed_path): 385 | extensions = ['.safetensors', '.pt', '.bin'] 386 | for x in extensions: 387 | t = embed_path + x 388 | if os.path.isfile(t): 389 | valid_file = t 390 | break 391 | else: 392 | valid_file = embed_path 393 | if valid_file is not None: 394 | break 395 | 396 | if valid_file is None: 397 | return None 398 | 399 | embed_path = valid_file 400 | 401 | embed_out = None 402 | 403 | try: 404 | if embed_path.lower().endswith(".safetensors"): 405 | import safetensors.torch 406 | embed = safetensors.torch.load_file(embed_path, device="cpu") 407 | else: 408 | if 'weights_only' in torch.load.__code__.co_varnames: 409 | try: 410 | embed = torch.load(embed_path, weights_only=True, map_location="cpu") 411 | except: 412 | embed_out = safe_load_embed_zip(embed_path) 413 | else: 414 | embed = torch.load(embed_path, map_location="cpu") 415 | except Exception as e: 416 | print(traceback.format_exc()) 417 | print() 418 | print("error loading embedding, skipping loading:", embedding_name) 419 | return None 420 | 421 | if embed_out is None: 422 | if 'string_to_param' in embed: 423 | values = embed['string_to_param'].values() 424 | embed_out = next(iter(values)) 425 | elif isinstance(embed, list): 426 | out_list = [] 427 | for x in range(len(embed)): 428 | for k in embed[x]: 429 | t = embed[x][k] 430 | if t.shape[-1] != embedding_size: 431 | continue 432 | out_list.append(t.reshape(-1, t.shape[-1])) 433 | embed_out = torch.cat(out_list, dim=0) 434 | elif embed_key is not None and embed_key in embed: 435 | embed_out = embed[embed_key] 436 | else: 437 | values = embed.values() 438 | embed_out = next(iter(values)) 439 | return embed_out 440 | 441 | class SDTokenizer: 442 | def __init__(self, tokenizer_path=None, max_length=77, pad_with_end=True, embedding_directory=None, embedding_size=768, embedding_key='clip_l', tokenizer_class=CLIPTokenizer, has_start_token=True, pad_to_max_length=True): 443 | if tokenizer_path is None: 444 | tokenizer_path = os.path.join(appropriate_file_path, "sd1_tokenizer") 445 | self.tokenizer = tokenizer_class.from_pretrained(tokenizer_path) 446 | self.max_length = max_length 447 | 448 | empty = self.tokenizer('')["input_ids"] 449 | if has_start_token: 450 | self.tokens_start = 1 451 | self.start_token = empty[0] 452 | self.end_token = empty[1] 453 | else: 454 | self.tokens_start = 0 455 | self.start_token = None 456 | self.end_token = empty[0] 457 | self.pad_with_end = pad_with_end 458 | self.pad_to_max_length = pad_to_max_length 459 | 460 | vocab = self.tokenizer.get_vocab() 461 | self.inv_vocab = {v: k for k, v in vocab.items()} 462 | self.embedding_directory = embedding_directory 463 | self.max_word_length = 8 464 | self.embedding_identifier = "embedding:" 465 | self.embedding_size = embedding_size 466 | self.embedding_key = embedding_key 467 | 468 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 469 | # This function is only for reference, and not used in the backend or runtime. 470 | def _try_get_embedding(self, embedding_name:str): 471 | ''' 472 | Takes a potential embedding name and tries to retrieve it. 473 | Returns a Tuple consisting of the embedding and any leftover string, embedding can be None. 474 | ''' 475 | embed = load_embed(embedding_name, self.embedding_directory, self.embedding_size, self.embedding_key) 476 | if embed is None: 477 | stripped = embedding_name.strip(',') 478 | if len(stripped) < len(embedding_name): 479 | embed = load_embed(stripped, self.embedding_directory, self.embedding_size, self.embedding_key) 480 | return (embed, embedding_name[len(stripped):]) 481 | return (embed, "") 482 | 483 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 484 | # This function is only for reference, and not used in the backend or runtime. 485 | def tokenize_with_weights(self, text:str, return_word_ids=False): 486 | ''' 487 | Takes a prompt and converts it to a list of (token, weight, word id) elements. 488 | Tokens can both be integer tokens and pre computed CLIP tensors. 489 | Word id values are unique per word and embedding, where the id 0 is reserved for non word tokens. 490 | Returned list has the dimensions NxM where M is the input size of CLIP 491 | ''' 492 | if self.pad_with_end: 493 | pad_token = self.end_token 494 | else: 495 | pad_token = 0 496 | 497 | text = escape_important(text) 498 | parsed_weights = token_weights(text, 1.0) 499 | 500 | #tokenize words 501 | tokens = [] 502 | for weighted_segment, weight in parsed_weights: 503 | to_tokenize = unescape_important(weighted_segment).replace("\n", " ").split(' ') 504 | to_tokenize = [x for x in to_tokenize if x != ""] 505 | for word in to_tokenize: 506 | #if we find an embedding, deal with the embedding 507 | if word.startswith(self.embedding_identifier) and self.embedding_directory is not None: 508 | embedding_name = word[len(self.embedding_identifier):].strip('\n') 509 | embed, leftover = self._try_get_embedding(embedding_name) 510 | if embed is None: 511 | print(f"warning, embedding:{embedding_name} does not exist, ignoring") 512 | else: 513 | if len(embed.shape) == 1: 514 | tokens.append([(embed, weight)]) 515 | else: 516 | tokens.append([(embed[x], weight) for x in range(embed.shape[0])]) 517 | #if we accidentally have leftover text, continue parsing using leftover, else move on to next word 518 | if leftover != "": 519 | word = leftover 520 | else: 521 | continue 522 | #parse word 523 | tokens.append([(t, weight) for t in self.tokenizer(word)["input_ids"][self.tokens_start:-1]]) 524 | 525 | #reshape token array to CLIP input size 526 | batched_tokens = [] 527 | batch = [] 528 | if self.start_token is not None: 529 | batch.append((self.start_token, 1.0, 0)) 530 | batched_tokens.append(batch) 531 | for i, t_group in enumerate(tokens): 532 | #determine if we're going to try and keep the tokens in a single batch 533 | is_large = len(t_group) >= self.max_word_length 534 | 535 | while len(t_group) > 0: 536 | if len(t_group) + len(batch) > self.max_length - 1: 537 | remaining_length = self.max_length - len(batch) - 1 538 | #break word in two and add end token 539 | if is_large: 540 | batch.extend([(t,w,i+1) for t,w in t_group[:remaining_length]]) 541 | batch.append((self.end_token, 1.0, 0)) 542 | t_group = t_group[remaining_length:] 543 | #add end token and pad 544 | else: 545 | batch.append((self.end_token, 1.0, 0)) 546 | if self.pad_to_max_length: 547 | batch.extend([(pad_token, 1.0, 0)] * (remaining_length)) 548 | #start new batch 549 | batch = [] 550 | if self.start_token is not None: 551 | batch.append((self.start_token, 1.0, 0)) 552 | batched_tokens.append(batch) 553 | else: 554 | batch.extend([(t,w,i+1) for t,w in t_group]) 555 | t_group = [] 556 | 557 | #fill last batch 558 | batch.append((self.end_token, 1.0, 0)) 559 | if self.pad_to_max_length: 560 | batch.extend([(pad_token, 1.0, 0)] * (self.max_length - len(batch))) 561 | 562 | if not return_word_ids: 563 | batched_tokens = [[(t, w) for t, w,_ in x] for x in batched_tokens] 564 | 565 | return batched_tokens 566 | 567 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 568 | # This function is only for reference, and not used in the backend or runtime. 569 | def untokenize(self, token_weight_pair): 570 | return list(map(lambda a: (a, self.inv_vocab[a[0]]), token_weight_pair)) 571 | 572 | 573 | class SD1Tokenizer: 574 | def __init__(self, embedding_directory=None, clip_name="l", tokenizer=SDTokenizer): 575 | self.clip_name = clip_name 576 | self.clip = "clip_{}".format(self.clip_name) 577 | setattr(self, self.clip, tokenizer(embedding_directory=embedding_directory)) 578 | 579 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 580 | # This function is only for reference, and not used in the backend or runtime. 581 | def tokenize_with_weights(self, text:str, return_word_ids=False): 582 | out = {} 583 | out[self.clip_name] = getattr(self, self.clip).tokenize_with_weights(text, return_word_ids) 584 | return out 585 | 586 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 587 | # This function is only for reference, and not used in the backend or runtime. 588 | def untokenize(self, token_weight_pair): 589 | return getattr(self, self.clip).untokenize(token_weight_pair) 590 | 591 | 592 | class SD1ClipModel(torch.nn.Module): 593 | def __init__(self, device="cpu", dtype=None, clip_name="l", clip_model=SDClipModel, **kwargs): 594 | super().__init__() 595 | self.clip_name = clip_name 596 | self.clip = "clip_{}".format(self.clip_name) 597 | setattr(self, self.clip, clip_model(device=device, dtype=dtype, **kwargs)) 598 | 599 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 600 | # This function is only for reference, and not used in the backend or runtime. 601 | def clip_layer(self, layer_idx): 602 | getattr(self, self.clip).clip_layer(layer_idx) 603 | 604 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 605 | # This function is only for reference, and not used in the backend or runtime. 606 | def reset_clip_layer(self): 607 | getattr(self, self.clip).reset_clip_layer() 608 | 609 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 610 | # This function is only for reference, and not used in the backend or runtime. 611 | def encode_token_weights(self, token_weight_pairs): 612 | token_weight_pairs = token_weight_pairs[self.clip_name] 613 | out, pooled = getattr(self, self.clip).encode_token_weights(token_weight_pairs) 614 | return out, pooled 615 | 616 | # Taken from https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/sd1_clip.py 617 | # This function is only for reference, and not used in the backend or runtime. 618 | def load_sd(self, sd): 619 | return getattr(self, self.clip).load_sd(sd) 620 | -------------------------------------------------------------------------------- /Line2Normalmap_modules/sd2_clip.py: -------------------------------------------------------------------------------- 1 | import sys 2 | # 'frozen' 状態に応じて適切なファイルパスを取得する関数 3 | def get_appropriate_file_path(): 4 | if getattr(sys, 'frozen', False): 5 | return sys.executable + "/Line2Normalmap/" 6 | else: 7 | return __file__ 8 | appropriate_file_path = get_appropriate_file_path() 9 | import sys 10 | # 'frozen'状態に応じて適切なファイルパスを取得する関数 11 | def get_appropriate_file_path(): 12 | if getattr(sys, 'frozen', False): 13 | # ビルドされたアプリケーションの場合、sys.executableのパスを使用 14 | return sys.executable + "/Line2Normalmap/" 15 | else: 16 | # そうでない場合は、従来通りappropriate_file_pathを使用 17 | return appropriate_file_path 18 | 19 | # 適切なファイルパスを取得 20 | appropriate_file_path = get_appropriate_file_path() 21 | 22 | import sys 23 | import os 24 | 25 | # 'frozen'状態に応じて適切なファイルパスを取得する関数 26 | def get_appropriate_file_path(): 27 | if getattr(sys, 'frozen', False): 28 | # ビルドされたアプリケーションの場合、os.path.dirname(sys.executable)のパスを使用 29 | return os.path.dirname(sys.executable) + "/ldm_patched/modules" 30 | else: 31 | # そうでない場合は、従来通りappropriate_file_pathを使用 32 | return os.path.dirname(appropriate_file_path) 33 | 34 | # 適切なファイルパスを取得 35 | appropriate_file_path = get_appropriate_file_path() 36 | 37 | # Taken from https://github.com/comfyanonymous/ComfyUI 38 | # This file is only for reference, and not used in the backend or runtime. 39 | 40 | 41 | from ldm_patched.modules import sd1_clip 42 | import torch 43 | import os 44 | 45 | class SD2ClipHModel(sd1_clip.SDClipModel): 46 | def __init__(self, arch="ViT-H-14", device="cpu", max_length=77, freeze=True, layer="penultimate", layer_idx=None, dtype=None): 47 | if layer == "penultimate": 48 | layer="hidden" 49 | layer_idx=-2 50 | 51 | textmodel_json_config = os.path.join(appropriate_file_path, "sd2_clip_config.json") 52 | super().__init__(device=device, freeze=freeze, layer=layer, layer_idx=layer_idx, textmodel_json_config=textmodel_json_config, dtype=dtype, special_tokens={"start": 49406, "end": 49407, "pad": 0}) 53 | 54 | class SD2ClipHTokenizer(sd1_clip.SDTokenizer): 55 | def __init__(self, tokenizer_path=None, embedding_directory=None): 56 | super().__init__(tokenizer_path, pad_with_end=False, embedding_directory=embedding_directory, embedding_size=1024) 57 | 58 | class SD2Tokenizer(sd1_clip.SD1Tokenizer): 59 | def __init__(self, embedding_directory=None): 60 | super().__init__(embedding_directory=embedding_directory, clip_name="h", tokenizer=SD2ClipHTokenizer) 61 | 62 | class SD2ClipModel(sd1_clip.SD1ClipModel): 63 | def __init__(self, device="cpu", dtype=None, **kwargs): 64 | super().__init__(device=device, dtype=dtype, clip_name="h", clip_model=SD2ClipHModel, **kwargs) 65 | -------------------------------------------------------------------------------- /Line2Normalmap_modules/sdxl_clip.py: -------------------------------------------------------------------------------- 1 | import sys 2 | # 'frozen' 状態に応じて適切なファイルパスを取得する関数 3 | def get_appropriate_file_path(): 4 | if getattr(sys, 'frozen', False): 5 | return sys.executable + "/Line2Normalmap/" 6 | else: 7 | return __file__ 8 | appropriate_file_path = get_appropriate_file_path() 9 | import sys 10 | # 'frozen'状態に応じて適切なファイルパスを取得する関数 11 | def get_appropriate_file_path(): 12 | if getattr(sys, 'frozen', False): 13 | # ビルドされたアプリケーションの場合、sys.executableのパスを使用 14 | return sys.executable + "/Line2Normalmap/" 15 | else: 16 | # そうでない場合は、従来通りappropriate_file_pathを使用 17 | return appropriate_file_path 18 | 19 | # 適切なファイルパスを取得 20 | appropriate_file_path = get_appropriate_file_path() 21 | 22 | import sys 23 | import os 24 | 25 | # 'frozen'状態に応じて適切なファイルパスを取得する関数 26 | def get_appropriate_file_path(): 27 | if getattr(sys, 'frozen', False): 28 | # ビルドされたアプリケーションの場合、os.path.dirname(sys.executable)のパスを使用 29 | return os.path.dirname(sys.executable) + "/ldm_patched/modules" 30 | else: 31 | # そうでない場合は、従来通りappropriate_file_pathを使用 32 | return os.path.dirname(appropriate_file_path) 33 | 34 | 35 | # 適切なファイルパスを取得 36 | appropriate_file_path = get_appropriate_file_path() 37 | 38 | 39 | # Taken from https://github.com/comfyanonymous/ComfyUI 40 | # This file is only for reference, and not used in the backend or runtime. 41 | 42 | 43 | from ldm_patched.modules import sd1_clip 44 | import torch 45 | import os 46 | 47 | class SDXLClipG(sd1_clip.SDClipModel): 48 | def __init__(self, device="cpu", max_length=77, freeze=True, layer="penultimate", layer_idx=None, dtype=None): 49 | if layer == "penultimate": 50 | layer="hidden" 51 | layer_idx=-2 52 | 53 | textmodel_json_config = os.path.join(appropriate_file_path, "clip_config_bigg.json") 54 | super().__init__(device=device, freeze=freeze, layer=layer, layer_idx=layer_idx, textmodel_json_config=textmodel_json_config, dtype=dtype, 55 | special_tokens={"start": 49406, "end": 49407, "pad": 0}, layer_norm_hidden_state=False) 56 | 57 | def load_sd(self, sd): 58 | return super().load_sd(sd) 59 | 60 | class SDXLClipGTokenizer(sd1_clip.SDTokenizer): 61 | def __init__(self, tokenizer_path=None, embedding_directory=None): 62 | super().__init__(tokenizer_path, pad_with_end=False, embedding_directory=embedding_directory, embedding_size=1280, embedding_key='clip_g') 63 | 64 | 65 | class SDXLTokenizer: 66 | def __init__(self, embedding_directory=None): 67 | self.clip_l = sd1_clip.SDTokenizer(embedding_directory=embedding_directory) 68 | self.clip_g = SDXLClipGTokenizer(embedding_directory=embedding_directory) 69 | 70 | def tokenize_with_weights(self, text:str, return_word_ids=False): 71 | out = {} 72 | out["g"] = self.clip_g.tokenize_with_weights(text, return_word_ids) 73 | out["l"] = self.clip_l.tokenize_with_weights(text, return_word_ids) 74 | return out 75 | 76 | def untokenize(self, token_weight_pair): 77 | return self.clip_g.untokenize(token_weight_pair) 78 | 79 | class SDXLClipModel(torch.nn.Module): 80 | def __init__(self, device="cpu", dtype=None): 81 | super().__init__() 82 | self.clip_l = sd1_clip.SDClipModel(layer="hidden", layer_idx=-2, device=device, dtype=dtype, layer_norm_hidden_state=False) 83 | self.clip_g = SDXLClipG(device=device, dtype=dtype) 84 | 85 | def clip_layer(self, layer_idx): 86 | self.clip_l.clip_layer(layer_idx) 87 | self.clip_g.clip_layer(layer_idx) 88 | 89 | def reset_clip_layer(self): 90 | self.clip_g.reset_clip_layer() 91 | self.clip_l.reset_clip_layer() 92 | 93 | def encode_token_weights(self, token_weight_pairs): 94 | token_weight_pairs_g = token_weight_pairs["g"] 95 | token_weight_pairs_l = token_weight_pairs["l"] 96 | g_out, g_pooled = self.clip_g.encode_token_weights(token_weight_pairs_g) 97 | l_out, l_pooled = self.clip_l.encode_token_weights(token_weight_pairs_l) 98 | return torch.cat([l_out, g_out], dim=-1), g_pooled 99 | 100 | def load_sd(self, sd): 101 | if "text_model.encoder.layers.30.mlp.fc1.weight" in sd: 102 | return self.clip_g.load_sd(sd) 103 | else: 104 | return self.clip_l.load_sd(sd) 105 | 106 | class SDXLRefinerClipModel(sd1_clip.SD1ClipModel): 107 | def __init__(self, device="cpu", dtype=None): 108 | super().__init__(device=device, dtype=dtype, clip_name="g", clip_model=SDXLClipG) 109 | -------------------------------------------------------------------------------- /Line2Normalmap_modules/shared_cmd_options.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/tori29umai0123/Line2Normalmap/cbb6a109eddcfe02225e1a2ef1278f6fd9b665e3/Line2Normalmap_modules/shared_cmd_options.py -------------------------------------------------------------------------------- /Line2Normalmap_modules/ui_extensions.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/tori29umai0123/Line2Normalmap/cbb6a109eddcfe02225e1a2ef1278f6fd9b665e3/Line2Normalmap_modules/ui_extensions.py -------------------------------------------------------------------------------- /Line2Normalmap_setup.py: -------------------------------------------------------------------------------- 1 | import os 2 | import re 3 | import shutil 4 | 5 | # 絶対パスで作業ディレクトリを指定 6 | script_path = os.path.abspath(__file__) 7 | target_directory = os.path.dirname(script_path) 8 | 9 | # 追加するコードの定義 10 | prepend_code = """import sys 11 | # 'frozen' 状態に応じて適切なファイルパスを取得する関数 12 | def get_appropriate_file_path(): 13 | if getattr(sys, 'frozen', False): 14 | return sys.executable + "/Line2Normalmap/" 15 | else: 16 | return __file__ 17 | appropriate_file_path = get_appropriate_file_path() 18 | """ 19 | 20 | ldm_special_prepend_code = """import sys 21 | import os 22 | def get_appropriate_file_path(): 23 | if getattr(sys, 'frozen', False): 24 | return os.path.dirname(sys.executable) + "/ldm_patched/modules" 25 | else: 26 | return os.path.dirname(__file__) 27 | appropriate_file_path = get_appropriate_file_path() 28 | """ 29 | 30 | exclude_files = [] 31 | 32 | ldm_patched_files = [ 33 | os.path.join(target_directory, "ldm_patched/modules/sd1_clip.py"), 34 | os.path.join(target_directory, "ldm_patched/modules/sd2_clip.py"), 35 | os.path.join(target_directory, "ldm_patched/modules/sdxl_clip.py"), 36 | ] 37 | 38 | exclude_folders = [ 39 | os.path.join(target_directory, "venv"), 40 | os.path.join(target_directory, "Line2Normalmap_modules"), 41 | os.path.join(target_directory, "utils"), 42 | ] 43 | 44 | # 新しいファイルの追加に加えて既存のコピー機能 45 | files_to_copy = { 46 | os.path.join(target_directory, "Line2Normalmap_modules/config_states.py"): os.path.join(target_directory, "modules/config_states.py"), 47 | os.path.join(target_directory, "Line2Normalmap_modules/gitpython_hack.py"): os.path.join(target_directory, "modules/gitpython_hack.py"), 48 | os.path.join(target_directory, "Line2Normalmap_modules/launch_utils_Line2Normalmap.py"): os.path.join(target_directory, "modules/launch_utils_Line2Normalmap.py"), 49 | os.path.join(target_directory, "Line2Normalmap_modules/sdxl_clip.py"): os.path.join(target_directory, "ldm_patched/modules/sdxl_clip.py"), 50 | os.path.join(target_directory, "Line2Normalmap_modules/sd1_clip.py"): os.path.join(target_directory, "ldm_patched/modules/sd1_clip.py"), 51 | os.path.join(target_directory, "Line2Normalmap_modules/sd2_clip.py"): os.path.join(target_directory, "ldm_patched/modules/sd2_clip.py"), 52 | os.path.join(target_directory, "Line2Normalmap_modules/shared_cmd_options.py"): os.path.join(target_directory, "modules/shared_cmd_options.py"), 53 | os.path.join(target_directory, "Line2Normalmap_modules/ui_extensions.py"): os.path.join(target_directory, "modules/ui_extensions.py"), 54 | } 55 | 56 | def file_needs_update(filepath): 57 | try: 58 | with open(filepath, 'r', encoding='utf-8') as file: 59 | content = file.read() 60 | return re.search(r'(? 9 | ②セキュリティーソフトの設定で、フォルダと実行ファイル名を除外リストに追加する。
10 | 例：Windows Defenderの場合、Windows セキュリティ→ウイルスと脅威の防止→ウイルスと脅威の防止の設定→設定の管理→除外
11 | Line2Normalmap.exe(プロセス)
12 | C:\Line2Normalmap（フォルダ）
13 | のように指定する。
14 | ③venv.cmdを実行。 15 | ``` 16 | pyinstaller "C:\Line2Normalmap\Line2Normalmap.py" ^ 17 | --clean ^ 18 | --collect-data tkinterdnd2 ^ 19 | --add-data "C:\Line2Normalmap\javascript;.\javascript" ^ 20 | --add-data "C:\Line2Normalmap\ldm_patched;.\ldm_patched" ^ 21 | --add-data "C:\Line2Normalmap\localizations;.\localizations" ^ 22 | --add-data "C:\Line2Normalmap\modules;.\modules" ^ 23 | --add-data "C:\Line2Normalmap\modules_forge;.\modules_forge" ^ 24 | --add-data "C:\Line2Normalmap\repositories;.\repositories" ^ 25 | --add-data "C:\Line2Normalmap\cache.json;." ^ 26 | --add-data "C:\Line2Normalmap\script.js;." ^ 27 | --add-data "C:\Line2Normalmap\ui-config.json;." ^ 28 | --add-data "C:\Line2Normalmap\config_states;.\config_states" ^ 29 | --add-data "C:\Line2Normalmap\configs;.\configs" ^ 30 | --add-data "C:\Line2Normalmap\extensions-builtin;.\extensions-builtin" ^ 31 | --add-data "C:\Line2Normalmap\html;.\html" 32 | 33 | xcopy /E /I /Y venv\Lib\site-packages\xformers dist\Line2Normalmap\_internal\xformers 34 | xcopy /E /I /Y venv\Lib\site-packages\pytorch_lightning dist\Line2Normalmap\_internal\pytorch_lightning 35 | xcopy /E /I /Y venv\Lib\site-packages\lightning_fabric dist\Line2Normalmap\_internal\lightning_fabric 36 | xcopy /E /I /Y venv\Lib\site-packages\gradio dist\Line2Normalmap\_internal\gradio 37 | xcopy /E /I /Y venv\Lib\site-packages\gradio_client dist\Line2Normalmap\_internal\gradio_client 38 | xcopy /E /I /Y venv\Lib\site-packages\kornia dist\Line2Normalmap\_internal\kornia 39 | xcopy /E /I /Y venv\Lib\site-packages\open_clip dist\Line2Normalmap\_internal\open_clip 40 | xcopy /E /I /Y venv\Lib\site-packages\jsonmerge dist\Line2Normalmap\_internal\jsonmerge 41 | xcopy /E /I /Y venv\Lib\site-packages\torchdiffeq dist\Line2Normalmap\_internal\torchdiffeq 42 | xcopy /E /I /Y venv\Lib\site-packages\cleanfid dist\Line2Normalmap\_internal\cleanfid 43 | xcopy /E /I /Y venv\Lib\site-packages\clip dist\Line2Normalmap\_internal\clip 44 | xcopy /E /I /Y venv\Lib\site-packages\resize_right dist\Line2Normalmap\_internal\resize_right 45 | xcopy /E /I /Y venv\Lib\site-packages\diffusers dist\Line2Normalmap\_internal\diffusers 46 | xcopy /E /I /Y venv\Lib\site-packages\onnx dist\Line2Normalmap\_internal\onnx 47 | xcopy /E /I /Y venv\Lib\site-packages\onnxruntime dist\Line2Normalmap\_internal\onnxruntime 48 | xcopy /E /I /Y venv\Lib\site-packages\scipy dist\Line2Normalmap\_internal\scipy 49 | xcopy /E /I /Y config_states dist\Line2Normalmap\config_states 50 | xcopy /E /I /Y configs dist\Line2Normalmap\configs 51 | xcopy /E /I /Y embeddings dist\Line2Normalmap\embeddings 52 | xcopy /E /I /Y extensions-builtin dist\Line2Normalmap\extensions-builtin 53 | xcopy /E /I /Y html dist\Line2Normalmap\html 54 | xcopy /E /I /Y javascript dist\Line2Normalmap\javascript 55 | xcopy /E /I /Y ldm_patched dist\Line2Normalmap\ldm_patched 56 | xcopy /E /I /Y localizations dist\Line2Normalmap\localizations 57 | xcopy /E /I /Y modules dist\Line2Normalmap\modules 58 | xcopy /E /I /Y modules_forge dist\Line2Normalmap\modules_forge 59 | xcopy /E /I /Y repositories dist\Line2Normalmap\repositories 60 | xcopy /E /I /Y scripts dist\Line2Normalmap\scripts 61 | copy script.js dist\Line2Normalmap\script.js 62 | copy Line2Normalmap_model_DL.cmd dist\Line2Normalmap\Line2Normalmap_model_DL.cmd 63 | copy Line2Normalmap_ReadMe.txt dist\Line2Normalmap\Line2Normalmap_ReadMe.txt 64 | ``` 65 | -------------------------------------------------------------------------------- /utils/application.py: -------------------------------------------------------------------------------- 1 | import tkinter as tk 2 | from tkinter import ttk 3 | from tkinterdnd2 import DND_FILES, TkinterDnD 4 | import asyncio 5 | import cv2 6 | from PIL import Image, ImageTk 7 | import numpy as np 8 | import os 9 | import sys 10 | import datetime 11 | from utils.tagger import modelLoad, analysis 12 | from utils.request_api import create_and_save_images, get_model, set_model, get_controlnet_model 13 | 14 | 15 | if getattr(sys, 'frozen', False): 16 | # PyInstaller でビルドされた場合 17 | dpath = os.path.dirname(sys.executable) 18 | else: 19 | # 通常の Python スクリプトとして実行された場合 20 | dpath = os.path.dirname(sys.argv[0]) 21 | 22 | model = None 23 | fastapi_url = None 24 | 25 | 26 | def canny_process(image_path, threshold1, threshold2): 27 | img_pil = Image.open(image_path).convert('RGBA') 28 | img = np.array(img_pil) # PIL画像をNumPy配列に変換 29 | 30 | alpha_channel = img[:, :, 3] 31 | # RGBチャンネルを取得 32 | rgb_channels = img[:, :, :3] 33 | 34 | # アルファチャンネルで透明部分を白にする 35 | # アルファチャンネルが0の部分は白に、それ以外は元の色を使う 36 | white_background = np.ones_like(rgb_channels, dtype=np.uint8) * 255 37 | # アルファチャンネルを基に背景を合成 38 | white_background = cv2.bitwise_not(white_background, mask=alpha_channel) 39 | background = cv2.bitwise_or(white_background, white_background, mask=alpha_channel) 40 | foreground = cv2.bitwise_and(rgb_channels, rgb_channels, mask=alpha_channel) 41 | combined = cv2.add(foreground, background) 42 | 43 | # RGBA形式からRGB形式に変換 44 | combined = cv2.cvtColor(combined, cv2.COLOR_RGBA2RGB) 45 | 46 | # グレースケール変換 47 | gray = cv2.cvtColor(combined, cv2.COLOR_RGB2GRAY) 48 | # Cannyエッジ検出 49 | edges = cv2.Canny(gray, threshold1, threshold2) 50 | 51 | return edges 52 | 53 | def resize_image_aspect_ratio(image, max_length=1200): 54 | # 元の画像サイズを取得 55 | original_width, original_height = image.size 56 | 57 | # アスペクト比を計算 58 | aspect_ratio = original_width / original_height 59 | 60 | # 長辺がmax_lengthになるように新しいサイズを計算 61 | if original_width > original_height: 62 | new_width = max_length 63 | new_height = round(max_length / aspect_ratio) 64 | else: 65 | new_height = max_length 66 | new_width = round(max_length * aspect_ratio) 67 | 68 | # 新しいサイズで画像をリサイズ 69 | resized_image = image.resize((new_width, new_height), Image.Resampling.LANCZOS) 70 | return resized_image 71 | 72 | 73 | class Application(TkinterDnD.Tk): 74 | def __init__(self, fastapi_url=None): 75 | super().__init__() 76 | self.fastapi_url = fastapi_url 77 | self.title("Line2Normalmap") 78 | self.geometry("600x600") 79 | self.image_path = None 80 | self.canny_pil = None 81 | self.tab_control = ttk.Notebook(self) 82 | self.line_input_tab = tk.Frame(self.tab_control) 83 | self.canny_input_tab = tk.Frame(self.tab_control) 84 | self.image_output_tab = tk.Frame(self.tab_control) 85 | self.tab_control.add(self.line_input_tab, text='①線画入力') 86 | self.tab_control.add(self.canny_input_tab, text='②canny入力') 87 | self.tab_control.add(self.image_output_tab, text='③画像出力') 88 | self.tab_control.pack(expand=1, fill="both") 89 | self.setup_line_input_tab() 90 | self.setup_canny_input_tab() 91 | self.setup_image_output_tab() 92 | self.setup_drag_and_drop() 93 | self.sd_model_names = None 94 | self.sd_current_model_name = None 95 | 96 | def setup_line_input_tab(self): 97 | self.prompt_label = tk.Label(self.line_input_tab, text="Prompt:") 98 | self.prompt_label.pack() 99 | self.prompt_text = tk.Text(self.line_input_tab, width=60, height=3, wrap=tk.WORD) 100 | self.prompt_text.pack() 101 | self.analyze_prompt_button = tk.Button(self.line_input_tab, text="Prompt解析", command=self.analyze_prompt) 102 | self.analyze_prompt_button.pack() 103 | 104 | def setup_canny_input_tab(self): 105 | self.threshold1_slider = tk.Scale(self.canny_input_tab, from_=0, to=255, label="Threshold1", orient="horizontal") 106 | self.threshold1_slider.set(20) 107 | self.threshold1_slider.pack() 108 | self.threshold2_slider = tk.Scale(self.canny_input_tab, from_=0, to=255, label="Threshold2", orient="horizontal") 109 | self.threshold2_slider.set(120) 110 | self.threshold2_slider.pack() 111 | tk.Button(self.canny_input_tab, text="Canny加工", command=self.apply_canny).pack() 112 | self.clear_canny_button = tk.Button(self.canny_input_tab, text="Canny画像消去", command=self.clear_canny) 113 | self.clear_canny_button.pack() 114 | 115 | def setup_image_output_tab(self): 116 | self.sd_model_names, self.sd_current_model_name = get_model(self.fastapi_url) 117 | self.model_label = tk.Label(self.image_output_tab, text="モデル選択:") 118 | self.model_label.pack() 119 | # モデル選択プルダウンメニューを作成 120 | self.model_variable = tk.StringVar(self.image_output_tab) 121 | self.model_dropdown = tk.OptionMenu(self.image_output_tab, self.model_variable, *self.sd_model_names) 122 | self.model_dropdown.pack() 123 | # プルダウンメニューの選択が変更された時に呼び出される関数を登録 124 | self.model_variable.trace("w", self.on_model_selected) 125 | 126 | self.prompt_label = tk.Label(self.image_output_tab, text="Prompt:") 127 | self.prompt_label.pack() 128 | self.prompt_entry = tk.Text(self.image_output_tab, width=60, height=3, wrap=tk.WORD) 129 | self.prompt_entry.pack() 130 | self.negative_prompt_label = tk.Label(self.image_output_tab, text="Negative Prompt:") 131 | self.negative_prompt_label.pack() 132 | self.negative_prompt_entry = tk.Text(self.image_output_tab, width=60, height=3, wrap=tk.WORD) 133 | self.negative_prompt_entry.pack() 134 | negative = "lowres, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, blurry" 135 | self.negative_prompt_entry.insert(tk.END, negative) 136 | self.lineart_fidelity_label = tk.Label(self.image_output_tab, text="線画忠実度（1.0-1.5）:") 137 | self.lineart_fidelity_label.pack() 138 | self.lineart_fidelity = tk.DoubleVar(value=1.0) 139 | self.lineart_fidelity_slider = tk.Scale(self.image_output_tab, from_=1.0, to=1.5, resolution=0.05, orient=tk.HORIZONTAL, variable=self.lineart_fidelity) 140 | self.lineart_fidelity_slider.pack() 141 | self.generate_image_button = tk.Button(self.image_output_tab, text="画像生成", command=self.generate_image) 142 | self.generate_image_button.pack() 143 | 144 | 145 | def on_model_selected(self, *args): 146 | self.sd_current_model_name = get_model(self.fastapi_url) 147 | self.selected_model = self.model_variable.get() 148 | 149 | def show_processed_image(self, img): 150 | # PILライブラリを使用してNumPy配列から画像を作成 151 | image = Image.fromarray(img) 152 | 153 | # アスペクト比を保ちつつ、長辺が400ピクセルになるようにリサイズする処理 154 | original_width, original_height = image.size 155 | max_length = 400 156 | 157 | # アスペクト比を計算 158 | aspect_ratio = original_width / original_height 159 | 160 | # 長辺が400ピクセルになるように新しいサイズを計算 161 | if original_width > original_height: 162 | new_width = max_length 163 | new_height = round(max_length / aspect_ratio) 164 | else: 165 | new_height = max_length 166 | new_width = round(max_length * aspect_ratio) 167 | 168 | # 新しいサイズで画像をリサイズ 169 | image_resized = image.resize((new_width, new_height), Image.Resampling.LANCZOS) 170 | 171 | # リサイズした画像をTkinter用のPhotoImageに変換 172 | photo = ImageTk.PhotoImage(image_resized) 173 | 174 | # 現在選択されているタブを取得 175 | current_tab = self.tab_control.nametowidget(self.tab_control.select()) 176 | 177 | # 以前に表示した画像があれば削除 178 | if hasattr(current_tab, 'image_label'): 179 | current_tab.image_label.pack_forget() 180 | current_tab.image_label.destroy() 181 | 182 | # 新しい画像を表示するLabelを作成して配置 183 | current_tab.image_label = tk.Label(current_tab, image=photo) 184 | current_tab.image_label.image = photo # 参照を保持しておかないと画像が表示されなくなる 185 | current_tab.image_label.pack() 186 | 187 | 188 | def clear_processed_image(self): 189 | current_tab = self.tab_control.nametowidget(self.tab_control.select()) 190 | if hasattr(current_tab, 'image_label'): 191 | current_tab.image_label.pack_forget() 192 | current_tab.image_label.destroy() 193 | delattr(current_tab, 'image_label') # この行を追加 194 | 195 | def setup_drag_and_drop(self): 196 | for tab in [self.line_input_tab, self.canny_input_tab]: 197 | tab.drop_target_register(DND_FILES) 198 | tab.dnd_bind('<>', self.on_drop) 199 | 200 | def on_drop(self, event): 201 | files = self.parse_dropped_files(event.data) 202 | if files: 203 | try: 204 | self.image_path = files[0].encode('utf-8').decode(sys.getfilesystemencoding()) 205 | except UnicodeEncodeError as e: 206 | print(f"Error processing file name: {e}") 207 | return 208 | self.load_image(self.image_path) 209 | 210 | def parse_dropped_files(self, data): 211 | files = data.split() 212 | return [file.replace('{', '').replace('}', '') for file in files] 213 | 214 | def load_image(self, image_path): 215 | self.image_path = image_path # 画像パスを更新する 216 | 217 | img = Image.open(self.image_path).convert("RGBA") 218 | canvas = Image.new('RGBA', img.size, (255, 255, 255, 255)) # 白背景のキャンバスを作成 219 | img = Image.alpha_composite(canvas, img) 220 | img = img.convert("RGB") # 最終的な画像をRGB形式に変換 221 | 222 | # アスペクト比を保ちつつ、長辺が400ピクセルになるようにリサイズ 223 | max_size = (400, 400) 224 | img.thumbnail(max_size, Image.Resampling.LANCZOS) 225 | 226 | # 長辺を400にするためのリサイズ処理 227 | original_size = img.size 228 | ratio = float(max_size[0]) / max(original_size) 229 | new_size = tuple([int(x * ratio) for x in original_size]) 230 | photo_img = img.resize(new_size, Image.Resampling.LANCZOS) 231 | 232 | photo = ImageTk.PhotoImage(photo_img) 233 | 234 | current_tab = self.tab_control.nametowidget(self.tab_control.select()) 235 | 236 | # 既存のimage_labelがある場合、新しい画像で更新する 237 | if hasattr(current_tab, 'image_label'): 238 | current_tab.image_label.configure(image=photo) 239 | current_tab.image_label.image = photo 240 | else: 241 | current_tab.image_label = tk.Label(current_tab, image=photo) 242 | current_tab.image_label.image = photo 243 | current_tab.image_label.pack() 244 | 245 | # 画像の種類に応じて適切な変数を更新 246 | if current_tab == self.canny_input_tab: 247 | self.canny_pil = img # Canny用PIL画像を更新 248 | 249 | def analyze_prompt(self): 250 | global model 251 | model_dir = os.path.join(dpath, 'models/tagger') 252 | if model is None: 253 | model = modelLoad(model_dir) 254 | image_path = self.image_path 255 | tag = analysis(image_path, model_dir, model) 256 | execute_tags = ["monochrome", "greyscale", "lineart", "white background"] 257 | tag_list = tag.split(", ") 258 | filtered_tags = [t for t in tag_list if t not in execute_tags] 259 | new_tag = ", ".join(filtered_tags) 260 | self.prompt_text.delete("1.0", tk.END) 261 | self.prompt_entry.delete("1.0", tk.END) 262 | self.prompt_text.insert("1.0", new_tag) 263 | self.prompt_entry.insert("1.0", new_tag) 264 | 265 | def apply_canny(self): 266 | if self.image_path is None: 267 | return 268 | threshold1 = self.threshold1_slider.get() 269 | threshold2 = self.threshold2_slider.get() 270 | canny = canny_process(self.image_path, threshold1, threshold2) 271 | self.canny_pil = Image.fromarray(cv2.cvtColor(canny, cv2.COLOR_GRAY2RGB)) # Canny結果をPIL画像として保存 272 | self.show_processed_image(canny) 273 | 274 | def clear_canny(self): 275 | self.clear_processed_image() 276 | self.canny_pil = None 277 | 278 | def generate_image(self): 279 | output_dir = os.path.join(dpath, "output/") 280 | if not os.path.exists(output_dir): 281 | os.makedirs(output_dir) 282 | # self.image_path.splitから拡張子なしの画像の名前部分を抽出 283 | img_name = os.path.splitext(os.path.basename(self.image_path))[0] 284 | # 日時の文字列からファイル名として無効な文字を置換 285 | dt_now = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S") 286 | output_path = os.path.join(output_dir, img_name + "_" + dt_now + ".png") 287 | prompt = "normal map, " + self.prompt_entry.get("1.0", tk.END).strip() 288 | nega = self.negative_prompt_entry.get("1.0", tk.END).strip() 289 | lineart_fidelity = float(self.lineart_fidelity_slider.get()) 290 | 291 | if self.canny_pil is None: 292 | self.canny_pil = Image.fromarray(cv2.cvtColor(canny_process(self.image_path, 20, 120), cv2.COLOR_GRAY2RGB)) 293 | self.canny_pil = resize_image_aspect_ratio(self.canny_pil, 1200) 294 | output_pil = create_and_save_images(self.fastapi_url, prompt, nega, self.canny_pil, lineart_fidelity, output_path) 295 | output_np = np.array(output_pil) 296 | self.show_processed_image(output_np) 297 | 298 | def start(fastapi_url): 299 | app = Application(fastapi_url) 300 | app.mainloop() -------------------------------------------------------------------------------- /utils/request_api.py: -------------------------------------------------------------------------------- 1 | import requests 2 | import json 3 | import base64 4 | from datetime import datetime 5 | import os 6 | import itertools 7 | import random 8 | import re 9 | from PIL import Image, PngImagePlugin, ImageEnhance, ImageFilter, ImageOps 10 | import io 11 | import glob 12 | import cv2 13 | 14 | 15 | def build_payload(prompt, nega, w, h, unit1): 16 | return { 17 | "prompt": prompt, 18 | "negative_prompt": nega, 19 | "seed": -1, 20 | "sampler_name": "Euler a", 21 | "steps": 20, 22 | "cfg_scale": 7, 23 | "width": w, 24 | "height": h, 25 | "alwayson_scripts": {"ControlNet": {"args": [unit1]}}, 26 | } 27 | 28 | def send_post_request(url, payload): 29 | headers = { 30 | "Content-Type": "application/json" 31 | } 32 | response = requests.post(url, data=json.dumps(payload), headers=headers) 33 | return response 34 | 35 | 36 | def save_image(data, url, file_name): 37 | image_string = data["images"][0] 38 | image_bytes = base64.b64decode(image_string) 39 | 40 | png_payload = { 41 | "image": "data:image/png;base64," + image_string 42 | } 43 | response2 = requests.post(url=f'{url}/sdapi/v1/png-info', json=png_payload) 44 | image_info = response2.json().get("info") 45 | 46 | image = Image.open(io.BytesIO(image_bytes)) 47 | pnginfo = PngImagePlugin.PngInfo() 48 | if image_info: # Ensure image_info is not None 49 | pnginfo.add_text("parameters", image_info) 50 | 51 | image.save(file_name, pnginfo=pnginfo) 52 | return image 53 | 54 | 55 | def create_and_save_images(input_url, prompt, nega, canny_pil, lineart_fidelity, output_path): 56 | url = f"{input_url}/sdapi/v1/txt2img" 57 | w, h = canny_pil.size 58 | canny_bytes = io.BytesIO() 59 | canny_pil.save(canny_bytes, format='PNG') 60 | encoded_canny = base64.b64encode(canny_bytes.getvalue()).decode('utf-8') 61 | 62 | prompt = "masterpiece, best quality, SimplepositiveXLv1 , " + prompt 63 | unit1 = { 64 | "image": encoded_canny, 65 | "mask_image": None, 66 | "control_mode": "Balanced", 67 | "enabled": True, 68 | "guidance_end": 1, 69 | "guidance_start": 0, 70 | "pixel_perfect": True, 71 | "processor_res": 1200, 72 | "resize_mode": "Just Resize", # "Just Resize", "Crop and Resize", "Resize and Fill" 73 | "threshold_a": 64, 74 | "threshold_b": 64, 75 | "weight": lineart_fidelity, 76 | "module": "canny", 77 | "model": "control-lora-canny-rank256 [ec2dbbe4]", 78 | "save_detected_map": None, 79 | "hr_option": "Both" 80 | } 81 | 82 | 83 | payload = build_payload(prompt, nega, w, h, unit1) 84 | response = send_post_request(url, payload) 85 | image_data = response.json() 86 | 87 | if "images" in image_data and image_data["images"]: 88 | output_pil = save_image(image_data, input_url, output_path) 89 | print(f"Downloaded {output_path} to local") 90 | return output_pil 91 | else: 92 | print("Failed to generate image. 'images' key not found in the response.") 93 | 94 | def get_model(url): 95 | sd_models = requests.get(f"{url}/sdapi/v1/sd-models").json() 96 | sd_model_names = [i["title"] for i in sd_models] 97 | current_model_name = requests.get(f"{url}/sdapi/v1/options").json()["sd_model_checkpoint"] 98 | return sd_model_names, current_model_name 99 | 100 | def get_controlnet_model(url): 101 | cn_models = requests.get(f"{url}/controlnet/model_list").json() 102 | return cn_models 103 | 104 | def set_model(url, sd_model_name): 105 | option_payload = { 106 | "sd_model_checkpoint":sd_model_name, 107 | } 108 | response = requests.post(url=f'{url}/sdapi/v1/options', json=option_payload) -------------------------------------------------------------------------------- /utils/tagger.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | # https://github.com/kohya-ss/sd-scripts/blob/main/finetune/tag_images_by_wd14_tagger.py 3 | 4 | import csv 5 | import os 6 | os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true' 7 | 8 | from PIL import Image 9 | import cv2 10 | import numpy as np 11 | from pathlib import Path 12 | import onnx 13 | import onnxruntime as ort 14 | 15 | # from wd14 tagger 16 | IMAGE_SIZE = 448 17 | 18 | model = None # Initialize model variable 19 | 20 | 21 | def convert_array_to_bgr(array): 22 | """ 23 | Convert a NumPy array image to BGR format regardless of its original format. 24 | 25 | Parameters: 26 | - array: NumPy array of the image. 27 | 28 | Returns: 29 | - A NumPy array representing the image in BGR format. 30 | """ 31 | # グレースケール画像（2次元配列） 32 | if array.ndim == 2: 33 | # グレースケールをBGRに変換（3チャンネルに拡張） 34 | bgr_array = np.stack((array,) * 3, axis=-1) 35 | # RGBAまたはRGB画像（3次元配列） 36 | elif array.ndim == 3: 37 | # RGBA画像の場合、アルファチャンネルを削除 38 | if array.shape[2] == 4: 39 | array = array[:, :, :3] 40 | # RGBをBGRに変換 41 | bgr_array = array[:, :, ::-1] 42 | else: 43 | raise ValueError("Unsupported array shape.") 44 | 45 | return bgr_array 46 | 47 | 48 | def preprocess_image(image): 49 | image = np.array(image) 50 | image = convert_array_to_bgr(image) 51 | 52 | size = max(image.shape[0:2]) 53 | pad_x = size - image.shape[1] 54 | pad_y = size - image.shape[0] 55 | pad_l = pad_x // 2 56 | pad_t = pad_y // 2 57 | image = np.pad(image, ((pad_t, pad_y - pad_t), (pad_l, pad_x - pad_l), (0, 0)), mode="constant", constant_values=255) 58 | 59 | interp = cv2.INTER_AREA if size > IMAGE_SIZE else cv2.INTER_LANCZOS4 60 | image = cv2.resize(image, (IMAGE_SIZE, IMAGE_SIZE), interpolation=interp) 61 | 62 | image = image.astype(np.float32) 63 | return image 64 | 65 | def modelLoad(model_dir): 66 | onnx_path = os.path.join(model_dir, "model.onnx") 67 | # 実行プロバイダーをCPUのみに指定 68 | providers = ['CPUExecutionProvider'] 69 | # InferenceSessionの作成時にプロバイダーのリストを指定 70 | ort_session = ort.InferenceSession(onnx_path, providers=providers) 71 | input_name = ort_session.get_inputs()[0].name 72 | 73 | # 実際に使用されているプロバイダーを取得して表示 74 | actual_provider = ort_session.get_providers()[0] # 使用されているプロバイダー 75 | print(f"Using provider: {actual_provider}") 76 | 77 | return [ort_session, input_name] 78 | 79 | def analysis(image_path, model_dir, model): 80 | ort_session = model[0] 81 | input_name = model[1] 82 | 83 | with open(os.path.join(model_dir, "selected_tags.csv"), "r", encoding="utf-8") as f: 84 | reader = csv.reader(f) 85 | l = [row for row in reader] 86 | header = l[0] # tag_id,name,category,count 87 | rows = l[1:] 88 | assert header[0] == "tag_id" and header[1] == "name" and header[2] == "category", f"unexpected csv format: {header}" 89 | 90 | general_tags = [row[1] for row in rows[1:] if row[2] == "0"] 91 | character_tags = [row[1] for row in rows[1:] if row[2] == "4"] 92 | 93 | tag_freq = {} 94 | undesired_tags = [] 95 | 96 | # 画像をロードして前処理する 97 | image_pil = Image.open(image_path).convert("RGB") 98 | 99 | image_preprocessed = preprocess_image(image_pil) 100 | image_preprocessed = np.expand_dims(image_preprocessed, axis=0) 101 | 102 | # 推論を実行 103 | prob = ort_session.run(None, {input_name: image_preprocessed})[0][0] 104 | # タグを生成 105 | combined_tags = [] 106 | general_tag_text = "" 107 | character_tag_text = "" 108 | remove_underscore = True 109 | caption_separator = ", " 110 | general_threshold = 0.35 111 | character_threshold = 0.35 112 | 113 | for i, p in enumerate(prob[4:]): 114 | if i < len(general_tags) and p >= general_threshold: 115 | tag_name = general_tags[i] 116 | if remove_underscore and len(tag_name) > 3: # ignore emoji tags like >_< and ^_^ 117 | tag_name = tag_name.replace("_", " ") 118 | 119 | if tag_name not in undesired_tags: 120 | tag_freq[tag_name] = tag_freq.get(tag_name, 0) + 1 121 | general_tag_text += caption_separator + tag_name 122 | combined_tags.append(tag_name) 123 | elif i >= len(general_tags) and p >= character_threshold: 124 | tag_name = character_tags[i - len(general_tags)] 125 | if remove_underscore and len(tag_name) > 3: 126 | tag_name = tag_name.replace("_", " ") 127 | 128 | if tag_name not in undesired_tags: 129 | tag_freq[tag_name] = tag_freq.get(tag_name, 0) + 1 130 | character_tag_text += caption_separator + tag_name 131 | combined_tags.append(tag_name) 132 | 133 | # 先頭のカンマを取る 134 | if len(general_tag_text) > 0: 135 | general_tag_text = general_tag_text[len(caption_separator) :] 136 | if len(character_tag_text) > 0: 137 | character_tag_text = character_tag_text[len(caption_separator) :] 138 | tag_text = caption_separator.join(combined_tags) 139 | return tag_text -------------------------------------------------------------------------------- /venv.cmd: -------------------------------------------------------------------------------- 1 | cmd /k venv\Scripts\activate --------------------------------------------------------------------------------