Subedit 字幕工作台
================
This deployment is a static, browser-only adaptation of:
https://github.com/hikashop-nicolas/subedit
Pinned source: 7ab6845367e04edf65c36bfb5a7c9fe3199edb2f (MIT).

Public source archive: ./source.zip
Dependency versions, resolved source URLs and integrity: ./DEPENDENCIES.json
License texts: ./licenses/

Local adaptations: Chinese demo/default locale and sample, portal return link;
Windows file URL handling; exact source-tarball dependency pins; browser-only
dependency preparation; isolation headers and public source/license packaging.
No login, server API, API key or paid service is required.

Pinned browser dependencies
--------------------------
mediabunny 1.51.0 (MPL-2.0)
https://github.com/hikashop-nicolas/mediabunny/tree/3ccd32106000306fb8d6ce23467cbd41e71d0195
Exact TypeScript source and MPL license are included in source.zip. The browser
ES module build is the committed upstream build; the source itself is unmodified.

mediaplay 0.1.0 (MIT)
https://github.com/hikashop-nicolas/mediaplay/tree/1b0c7fd72f958e0d22c4f040b981ccbab72f7390

localml 0.1.0 (MIT)
https://github.com/hikashop-nicolas/localml/tree/b44965328a571b09e563b969f50bff81abd91ff6
Compiled locally from the pinned source with TypeScript. Worker URL extensions
are adjusted from .ts to .js in generated files for the Vite production build.

Optional media decoding / rendering libraries
---------------------------------------------
libav.js 6.9.8.1 audio build incorporates FFmpeg 8.1 (LGPL-2.1-or-later).
The three separately replaceable libav assets are copied unchanged from the
pinned mediaplay source. They include their complete LGPL notice in the glue JS.
Build configuration and instructions: ./libav/NOTICE.md
Matching libav.js version source:
https://github.com/Yahweasel/libav.js/tree/65ce461de1add4c96bc74fac52bd239f900ec11f
https://github.com/Yahweasel/libav.js/archive/65ce461de1add4c96bc74fac52bd239f900ec11f.tar.gz
Matching FFmpeg source: https://ffmpeg.org/releases/ffmpeg-8.1.tar.xz
The custom audio configuration is recorded verbatim in libav/NOTICE.md; the
versioned libav.js Makefile supplies dependency versions and the build scripts.

@jellyfin/libass-wasm 4.2.4 / JavascriptSubtitlesOctopus
Copyrights and component license texts: ./octopus/COPYRIGHT
Published package gitHead / matching complete source checkout:
https://github.com/jellyfin/JavascriptSubtitlesOctopus/tree/2d4ec9b16a639f23733cd32982e09e0eb757f092
Clone recursively so its pinned libass, FriBidi, FreeType, Fontconfig, HarfBuzz,
Expat and Brotli submodules are included; run the upstream Docker build script.
The WASM, worker and fallback font are copied unchanged from the npm release.

Optional on-device AI
---------------------
@huggingface/transformers 4.2.0 and tokenizers (Apache-2.0), ONNX Runtime (MIT).
ONNX Runtime Web version: 1.26.0-dev.20260416-b7804b056c
Matching source: https://github.com/microsoft/onnxruntime/tree/b7804b056c30aa35c1748f8e4e239d0e2ff25d6d
Speech recognition and translation download model files only when requested and
after the app's model-host consent. The model licenses belong to each selected
model. These optional downloads need network access, storage and sufficient
device memory; ordinary subtitle editing does not require a model download.

These open-source libraries remain under their respective licenses. Users may
replace the separate media decoder/renderer assets with compatible rebuilt ones.
This distribution provides no warranty; see the actual license texts.