docs: add model card

7523ad7 verified 1 day ago

2.45 kB

license: mit
language:
  - ar
  - bg
  - cs
  - da
  - de
  - el
  - en
  - es
  - fi
  - fr
  - he
  - hi
  - hu
  - it
  - ja
  - ko
  - ms
  - nl
  - 'no'
  - pl
  - pt
  - ro
  - ru
  - sk
  - sv
  - sw
  - ta
  - tr
  - vi
  - zh
library_name: phoonnx
pipeline_tag: text-to-speech
tags:
  - tts
  - chatterbox
  - onnx
  - phoonnx
  - offline
  - voice-cloning
  - codec-lm
  - multilingual

phoonnx-chatterbox-multilingual

Mirror of the multilingual variant of Chatterbox by Resemble AI, exported to ONNX for phoonnx, OpenVoiceOS's offline TTS library, covering 30 languages from Arabic to Vietnamese. Like the base model, this is an autoregressive codec-LM (Llama backbone) with zero-shot cloning from an untranscribed reference clip and an exaggeration control. See docs/training/engines/chatterbox.md and docs/cloning.md.

The multilingual variant uses its own ChatterboxMTLTokenizer. Japanese and Chinese script normalization need pip install phoonnx[chatterbox-multilingual] (pulls in pykakasi and spacy-pkuseg); Korean normalization is pure-Python and always available; Hebrew and Russian degrade gracefully when their optional stressors/diacritizers are absent.

Related repositories: phoonnx-chatterbox (base English), phoonnx-chatterbox-turbo (faster GPT-2-based variant), phoonnx-chatterbox-lahgtna (Arabic dialects, fine-tuned from this model).

Usage

Cloning support requires pip install phoonnx[cloning].

import wave
from phoonnx.model_manager import TTSModelManager

manager = TTSModelManager()
manager.load()
manager.merge_default_voices()

voice = manager.voices["chatterbox/multilingual/pt"].load()
with wave.open("out.wav", "wb") as wav_file:
    voice.synthesize_wav(
        "Olá, isto é um teste.",
        wav_file,
        speaker_reference="reference.wav",  # optional: clone this speaker
    )

License

Resemble AI released Chatterbox Multilingual under the MIT license (see onnx-community/chatterbox-multilingual-ONNX). phoonnx itself is Apache-2.0.