The coloring page maker's line-drawing models (lines-v1 and lines-bold-v1).

lines-v1: the default "Fine pencil" model.

Informative Drawings: Learning to Generate Line Drawings that Convey Geometry and Semantics
(Caroline Chan, Fredo Durand, Phillip Isola, CVPR 2022), "anime_style" weights,
https://github.com/carolineec/informative-drawings
Converted to ONNX, with the weights stored as float16 (lines-v1.w16.bin) and expanded to
float32 in the browser. Nothing else is changed.

MIT License

Copyright (c) 2022 Caroline Chan

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

lines-bold-v1: the "Bold outline" model, downloaded only when a person picks Bold outline.
The same network and the same authors' "contour_style" weights (Informative Drawings, Caroline Chan,
Fredo Durand, Phillip Isola, CVPR 2022, https://github.com/carolineec/informative-drawings), under the
same MIT licence and copyright notice as above. It reuses lines-v1.graph.onnx; only the weights
(lines-bold-v1.w16.bin, float16, the same layout as lines-v1.w16.bin) differ. Nothing else is changed.

birefnet-lite-512-v1: the background remover's "Better cutout" model, downloaded only when a person asks for it.
BiRefNet_lite, from "Bilateral Reference for High-Resolution Dichotomous Image Segmentation" (Peng Zheng, Dehong Gao,
Deng-Ping Fan, Li Liu, Jorma Laaksonen, Wanli Ouyang, Nicu Sebe; CAAI Artificial Intelligence Research, 2024),
https://github.com/ZhengPeng7/BiRefNet, MIT licence, Copyright (c) 2024 ZhengPeng. The 512 x 512 ONNX export is by
Studio Ludens (https://huggingface.co/studioludens/birefnet-lite-512, MIT) and is served here unchanged. The full
licence text is in LICENSE-birefnet-lite.txt, and the file's size and hash are in birefnet-lite-512-v1.manifest.json.

The "Upscale an image for print" tool's picture-enlarging models. The tool's page credits whichever model made a picture, and the full texts are beside this file.

Standard (realesr-general-x4v3-dni70-fp32.onnx, 4.9 MB, the default): Real-ESRGAN, Xintao Wang, Liangbin Xie, Chao Dong and Ying Shan,
https://github.com/xinntao/Real-ESRGAN, realesr-general-x4v3 at denoise strength 0.7, BSD-3-Clause. See LICENSE-realesr-general-x4v3.txt.

Fast (lsdircompact-v2-fp32.onnx, 2.5 MB, downloaded only when a person picks Fast): 4xLSDIRCompact v2 by Philip Hofmann (Phhofm),
https://github.com/Phhofm/models/releases/tag/4xLSDIRCompact2, licensed under Creative Commons Attribution 4.0 International
(CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). Changes: converted from the author's PyTorch .pth to ONNX so it runs in the
browser; the weights are the author's, with their float16 values written as float32 exactly. The author does not endorse this site.
The network architecture is Real-ESRGAN's SRVGGNetCompact (BSD-3-Clause). See LICENSE-lsdircompact-v2.txt and lsdircompact-v2.manifest.json.

Best quality (realesrgan-x4plus-fp32.onnx, 67 MB, downloaded only when a person picks Best quality, WebGPU only): Real-ESRGAN, RealESRGAN_x4plus,
Xintao Wang, Liangbin Xie, Chao Dong and Ying Shan, https://github.com/xinntao/Real-ESRGAN, BSD-3-Clause. Changes: converted from the PyTorch .pth to ONNX
(float32); the weights are unchanged. See LICENSE-realesrgan-x4plus.txt and realesrgan-x4plus.manifest.json.

slimsam-77-v1: the background remover's "Tap what to keep" model, downloaded only when a person first taps.
SlimSAM-77-uniform, from "SlimSAM: 0.1% Data Makes Segment Anything Slim" (Zigeng Chen, Gongfan Fang, Xinyin Ma, Xinchao Wang),
https://github.com/czg1225/SlimSAM, Apache License 2.0, a pruned version of Segment Anything (Alexander Kirillov et al., Meta AI,
https://github.com/facebookresearch/segment-anything, Apache License 2.0). The ONNX export is Xenova's
(https://huggingface.co/Xenova/slimsam-77-uniform, Apache License 2.0) and is served here unchanged (slimsam-77-encoder-v1.onnx and
slimsam-77-decoder-v1.onnx). The licence text is in LICENSE-slimsam.txt, and the files' sizes and hashes are in slimsam-77-v1.manifest.json.