ComfyUI custom nodes for Chatterbox TTS with multilingual support (23 languages).
This is the only ComfyUI Chatterbox implementation that supports the multilingual model!
- Multilingual TTS: Support for 23 languages with language-specific text processing
- Standard TTS: High-quality English text-to-speech
- Voice Conversion: Convert voice timbre while preserving speech content
- Voice Cloning: Clone voices from reference audio prompts
- ComfyUI Integration: Full integration with ComfyUI's model management system
| Code | Language | Code | Language | Code | Language |
|---|---|---|---|---|---|
| ar | Arabic | he | Hebrew | pl | Polish |
| da | Danish | hi | Hindi | pt | Portuguese |
| de | German | it | Italian | ru | Russian |
| el | Greek | ja | Japanese | sv | Swedish |
| en | English | ko | Korean | sw | Swahili |
| es | Spanish | ms | Malay | tr | Turkish |
| fi | Finnish | nl | Dutch | zh | Chinese |
| fr | French | no | Norwegian |
Generate speech in 23 languages with voice cloning support.
Inputs:
text: Text to synthesize (max 300 characters)language: Target language selectionexaggeration: Emotional expressiveness (0.25-2.0)temperature: Sampling randomness (0.05-5.0)cfg_weight: CFG guidance weight (0.0-1.0, set to 0 for language transfer)audio_prompt: Optional reference audio for voice cloning
Standard English TTS with high quality output.
Convert the voice in an audio file to match a target voice.
Search for "ComfyUI-For-ChatterBox" in ComfyUI Manager and install.
cd ComfyUI/custom_nodes
git clone https://github.com/your-repo/ComfyUI-For-ChatterBox.git
cd ComfyUI-For-ChatterBox
pip install -r requirements.txtModels are automatically downloaded from HuggingFace on first use:
- Standard TTS:
models/tts/chatterbox/resembleai_default_voice/ - Multilingual TTS:
models/tts/chatterbox/resembleai_multilingual/
When using a reference audio in a different language than the target:
- Set
cfg_weightto 0 to mitigate accent transfer - Use a reference audio that matches the target language for best quality
Japanese text is automatically converted to hiragana (kanji → hiragana, katakana preserved).
Chinese characters are converted to Cangjie codes for tokenization.
Korean syllables are decomposed into Jamo for proper pronunciation.
resemble-perth: Audio watermarking supportdicta_onnx: Hebrew diacritizationrussian_text_stresser: Russian stress marking
- Chatterbox by Resemble AI
- ComfyUI community
MIT License