Sources & licenses
Subtitle Studio credits & licenses
Subtitle Studio was built for LoschaZ. Speech recognition, translation and video encoding run in your browser with open-source software, open models and open fonts:
What we built
The editor, the caption timing and layout, the renderer shared by preview and export, the timeline, the styles, the translation splitting and the SRT, VTT, ASS, LRC and text formats were written for LoschaZ.
Components & licenses
- Transformers.js 4.3.1 by Hugging Face – Apache License 2.0 (license). Served from this website.
- Mediabunny 1.61.3 by Vanilagy and contributors – Mozilla Public License 2.0 (license, source). Served from this website, unchanged.
- ONNX Runtime Web by Microsoft – MIT License. Its WebAssembly engine is loaded from cdn.jsdelivr.net.
- Whisper speech recognition models by OpenAI – MIT License; the ONNX versions with word timing (base, small) are loaded from Hugging Face.
- Opus-MT translation models by the University of Helsinki (Helsinki-NLP) – CC BY 4.0; the browser versions by Xenova are loaded from Hugging Face only when a translation needs them.
- Caption fonts under the SIL Open Font License 1.1: Anton (license), Lilita One (license), Montserrat (license) and Playfair Display (license), served from this website; Inter and Outfit are the site’s own fonts.
Our changes
The libraries, models and fonts are used unchanged; the fonts are split into language subsets so a page only loads the letters it needs.