videopython-chatterbox
======================

This distribution is a fork of `chatterbox-tts` 0.1.7 by Resemble AI:

    https://github.com/resemble-ai/chatterbox
    https://pypi.org/project/chatterbox-tts/

The Python source under `src/chatterbox/` is redistributed unchanged from
upstream 0.1.7 with two small changes. First, in `src/chatterbox/__init__.py`:

    -__version__ = version("chatterbox-tts")
    +__version__ = version("videopython-chatterbox")

Upstream looks up its own version from installed distribution metadata by name.
Because this distribution is named `videopython-chatterbox`, that lookup raises
`PackageNotFoundError` at import time and the package is unimportable without
the change.

Second, in `src/chatterbox/models/t3/inference/alignment_stream_analyzer.py`,
alignment-repetition detection is gated on S > 5. With five or fewer text
tokens there are no earlier tokens to reduce, and the upstream expression
raises an empty-reduction IndexError. Long-tail and generated-token repetition
checks remain active. Other runtime source files are unchanged.

The fork also changes dependency metadata and adds regression tests (see
README.md for the rationale and verification).

The original work is licensed under the MIT License, Copyright (c) 2025
Resemble AI. The full license text is preserved verbatim in LICENSE and applies
unchanged to this distribution.

Model weights are NOT redistributed here. They are downloaded at runtime from
the Hugging Face Hub by upstream's own code, under whatever terms Resemble AI
publishes them.
