Skip to content

Add multilingual (French) TTS UI + fix installability#62

Open
Fahtialalaina wants to merge 3 commits into
petermg:mainfrom
Fahtialalaina:claude/elastic-roentgen-ce9d98
Open

Add multilingual (French) TTS UI + fix installability#62
Fahtialalaina wants to merge 3 commits into
petermg:mainfrom
Fahtialalaina:claude/elastic-roentgen-ce9d98

Conversation

@Fahtialalaina

Copy link
Copy Markdown

What changed

New feature — multilingual TTS UI

  • multilingual/app_ml.py: a standalone Gradio interface built on ChatterboxMultilingualTTS (from the official chatterbox-tts package) that supports French + 22 other languages.
    • Language dropdown (French preselected), optional reference-audio voice cloning, and exaggeration / CFG-weight / temperature / seed controls.
    • Device auto-select: CUDA → Apple Silicon MPS → CPU.
    • Intentionally isolated from the vendored English package (chatterbox.src.chatterbox) to avoid the top-level chatterbox import-name collision; meant to run from its own directory / virtualenv.

Installability fixes (universally beneficial)

  • requirements.txt:
    • Unpin auto-editor — the pinned 27.1.1 was removed from PyPI, so fresh installs failed to resolve. auto-editor is only invoked as an optional CLI subprocess, and the flags used are stable across versions.
    • Add pydub — it is imported by Chatter.py but was missing from requirements.
    • The CUDA --extra-index-url and pinned torch/torchaudio are left untouched.

Housekeeping

  • .gitignore: ignore the extra virtualenv, generated audio (*.wav, multilingual/outputs/) and the runtime-generated settings.json.

Notes for reviewers

  • The multilingual UI is additive and does not touch Chatter.py or the vendored English model path, so existing behavior is unchanged.
  • Tested end-to-end on Apple Silicon (M1 Pro, MPS): French generation works and produces valid audio.

🤖 Generated with Claude Code

Fahtialalaina and others added 3 commits July 1, 2026 08:30
- multilingual/app_ml.py: new Gradio UI using ChatterboxMultilingualTTS
  (chatterbox-tts) for French + 22 other languages, with optional voice
  cloning, exaggeration/CFG/temperature controls and seeding. Runs on
  CUDA / Apple Silicon MPS / CPU. Isolated from the vendored English
  package to avoid the `chatterbox` name collision.
- requirements.txt: unpin auto-editor (27.1.1 was removed from PyPI and
  broke fresh installs) and add pydub (imported by Chatter.py but missing
  from requirements).
- .gitignore: ignore extra venv, generated audio and runtime settings.json.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
multilingual/README.md: prerequisites, install, launch, usage,
supported languages, French tips, and troubleshooting.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
LOCAL_SETUP.md: prerequisites, macOS/CUDA install paths, launch,
settings.json overview, and troubleshooting for the main app.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant