OpenDubStage
Copyright 2026 The OpenDubStage contributors

This product is licensed under the Apache License, Version 2.0 (see LICENSE).

Origin of the code
------------------
All source files were written for this project by its contributors, or for the
earlier "stillfoley" project by the same contributors (opendubstage/foley,
tests/foley, docs/foley-*.md, scripts/), which was also Apache-2.0. No
third-party source code is bundled.

Model weights and optional backends
-----------------------------------
This repository does not bundle any model weights, datasets, or third-party
source code. Optional backends download and run third-party software and
weights that carry their own licences, which are NOT Apache-2.0 and are
frequently non-commercial:

- Woosh (Sony AI): code MIT / Apache-2.0; weights CC-BY-NC 4.0.
- Synchformer weights (downloaded by Woosh from hkchengrex/MMAudio): CC-BY-NC 4.0.
- roberta-large tokenizer/config (Hugging Face): MIT.

Reference data
--------------
opendubstage/restore/memofix.py contains the 1/3-octave long-term average
speech spectrum values published in Byrne et al. (1994), J. Acoust. Soc. Am.
96(4), "An international comparison of long-term average speech spectra",
used as numeric reference data with citation.

Example media
-------------
The photographs in examples/foley/public_domain/ are in the public domain;
see examples/foley/public_domain/SOURCES.md for provenance. All audio used by
the test suite is generated synthetically at test time.
