PhotoS — Licensing Notice
=========================

Copyright (c) 2026 PhotoS Contributors

This file summarizes the license of every artifact shipped by this project.
It is a convenience summary, NOT a substitute for the full license texts.

1. Source code — MIT
--------------------
All PhotoS source code is licensed under the MIT License (see LICENSE):

  * photo-s-tools            (core package: engine, CLI, REST, MCP, GUI)
  * photo-s-plugin-scunet    (plugin code)
  * photo-s-plugin-lut       (plugin code)
  * photo-s-plugin-auto-tone (plugin code)
  * tools/, tests/, packaging/, skills/

You may use, modify, and redistribute the code commercially.

2. Model weights — mixed
------------------------
Weights are NOT bundled in the wheels; they are downloaded on first use and
verified against a pinned sha256.

  * SCUNet denoising weights ................ MIT (commercial use permitted)
  * auto-tone weights ....................... CC-BY-NC 4.0 (see below)

"auto-tone weights" means every model file published on this repository's
GitHub Releases and on the ModelScope repository
(https://modelscope.cn/models/dwphoto/photo-s-auto-tone-v2), including:

  * auto_tone_v7_clean.pt
  * auto_tone_siglip_h192_d03.pt
  * clip_train_rag.npz
  * hand_features.npz
  * aesthetic_head.pt
  * lora_aesthetic.safetensors / lora_aesthetic_config.json
  * lora_advisor.safetensors   / lora_advisor_config.json

These are licensed under Creative Commons Attribution-NonCommercial 4.0
International (CC-BY-NC 4.0), because they were trained on the author's
personal Lightroom edit history.

  * Free (no permission needed): personal, hobby, educational, research, and
    other non-commercial use, including publishing your own edited photos.
  * Commercial use requires a separate license. Contact:
    <1634103640@qq.com> or <https://dwphoto.top/message>

See docs/COMMERCIAL.md for the full boundary table and FAQ.

3. Upstream model components
----------------------------
The auto-tone plugin builds on upstream vision-language models that are
fetched at runtime (not redistributed here). Their licenses allow commercial
use and derivative weights:

  * SigLIP  — Apache-2.0
  * CLIP    — MIT
  * Qwen3-VL-2B — Apache-2.0

The non-commercial restriction comes solely from the author's own trained
grading heads/LoRAs; the rights chain is single-origin and clean.

4. Third-party dependencies
---------------------------
PhotoS depends on third-party packages (Pillow, rawpy/libraw, numpy, and the
optional opencv-python-headless, piexif, watchdog, tifffile, pillow-heif,
pillow-avif-plugin, tkinterdnd2, onnxruntime, torch, open_clip_torch,
transformers). Each is governed by its own license. Consult each project
before redistribution.

5. PhotoS output
----------------
Your photos are yours. Nothing in any PhotoS license claims rights over your
input images, your EXIF metadata, or the images you produce. The auto-tone
weight license covers the use of the model as a tool only, and does not
extend to your work product.

Questions about licensing: <1634103640@qq.com>
