Text-to-Speech
Transformers.js
ONNX
Chatterbox
speech
speech-generation
voice-cloning
multilingual-tts
quantized
q4
Instructions to use BricksDisplay/chatterbox-multilingual-ONNX-q4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use BricksDisplay/chatterbox-multilingual-ONNX-q4 with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('text-to-speech', 'BricksDisplay/chatterbox-multilingual-ONNX-q4');
Commit History
-
Initial: q4 single-file fork (from ipsilondev) with config.json + tokenizer.json fixes
Verified signature- Forked from ipsilondev/chatterbox-multilingual-ONNX-q4 (q4-quantized, single-file ONNX, no .onnx_data sibling) - Added missing config.json (synthesized from English model + vocab_size=2454; use_external_data_format=false) - Patched tokenizer.json — added BOS/EOS/START_SPEECH/STOP_SPEECH/EXAGGERATION special tokens with content names matching the post_processor template (transformers.js's added_tokens_map[name] lookup needs the raw 'BOS'/'EOS'/etc names to resolve, otherwise post-processor positions silently become UNK) These mirror the fixes applied earlier to BricksDisplay/chatterbox-multilingual-ONNX, but for the q4 single-file variant.171d2d6