Imgy by AzenTool — third-party components

Runtime code is installed through npm; exact versions and full component license files are retained in the source package lock and node_modules package distributions. No competitor logos or templates are used.

React / React DOM: MIT, Meta Platforms, Inc. https://github.com/facebook/react
Vite: MIT, Evan You and contributors. https://github.com/vitejs/vite
Lucide icons: ISC, Lucide contributors. https://github.com/lucide-icons/lucide
JSZip: MIT, Stuart Knightley and contributors. https://github.com/Stuk/jszip
pdf-lib: MIT, Andrew Dillon. https://github.com/Hopding/pdf-lib
gifenc: MIT, Matt DesLauriers. https://github.com/mattdesl/gifenc
UTIF.js: MIT, Photopea. https://github.com/photopea/UTIF.js
ag-psd: MIT, Agamnentzar. https://github.com/Agamnentzar/ag-psd
heic-to 1.6.6: LGPL-3.0, Hopper Gee and contributors. https://github.com/hoppergee/heic-to
libheif (heic-to bundled decoder): LGPL-3.0, Dirk Farin and contributors. https://github.com/strukturag/libheif (build information: https://github.com/hoppergee/heic-to). Upstream source is available for replacement/rebuilding through these links.
html2canvas: MIT, Niklas von Hertzen. https://github.com/niklasvh/html2canvas
ONNX Runtime Web: MIT, Microsoft Corporation. https://github.com/microsoft/onnxruntime
TensorFlow.js: Apache-2.0, Google and contributors. https://github.com/tensorflow/tfjs
UpscalerJS + ESRGAN Slim: MIT, Kevin Scott. https://github.com/thekevinscott/UpscalerJS
Noto Sans: SIL Open Font License 1.1, Google. Full license: /imgy/fonts/OFL.txt

Background model: U²Netp, Apache-2.0. Original: https://github.com/xuebinqin/U-2-Net
Model converted by rembg maintainers; downloaded from https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2netp.onnx
Expected MD5 8e83ca70e441ab06c318d82300c84806. Model inference/preprocessing in Imgy is original TypeScript based on the documented rembg session normalization.

Face model: UltraFace RFB-320, MIT. Original: https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB
ONNX conversion/model provenance: https://github.com/onnx/models/tree/main/validated/vision/body_analysis/ultraface

Models and inference assets are served from AzenTool. Visitors' image pixels are not sent to the original model providers.

Portrait matting upgrade (October 2026): MODNet photographic portrait matting, Apache-2.0, original code and weights: https://github.com/ZHKKKe/MODNet
ONNX conversion: https://huggingface.co/Xenova/modnet (FP32 ONNX model, SHA-256 07c308cf0fc7e6e8b2065a12ed7fc07e1de8febb7dc7839d7b7f15dd66584df9). Original full license is in /imgy/licenses/modnet-Apache-2.0.txt. Portrait and detailed/quick general-subject model options are processed locally. No image is sent to model publishers.

Detailed general-subject model: full U²Net, Apache-2.0, original https://github.com/xuebinqin/U-2-Net ; ONNX distribution https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnx . Shares the U²Net license in /imgy/licenses/u2net-Apache-2.0.txt.