Chatterbox GGUF

--- license: mit language: - ar - da - de - el - en - es - fi - fr - he - hi - it - ja - ko - ms - nl - no - pl - pt - ru - sv - sw - tr - zh base_model: - ResembleAI/chatterbox pipeline_tag: text-to-speech tags: - tts - text-to-speech - chatterbox - flow-matching - hifi-gan - gguf - crispasr library_name: ggml --- # Chatterbox TTS — GGUF (ggml-quantised) GGUF / ggml conversion of [`ResembleAI/chatterbox`](https://huggingface.co/ResembleAI/chatterbox) for use with **[CrispStrobe/CrispASR](https://github.com/CrispStrobe/CrispASR)**. Chatterbox is a full TTS pipeline: character tokenizer → T3 (30-layer Llama AR, 520M) → speech tokens → S3Gen (Conformer encoder + UNet1D CFM denoiser, 10 Euler steps) → HiFTGenerator vocoder (conv chains + Snake activations + iSTFT) → 24 kHz WAV. Distributed under **MIT license**. This is the **multilingual** Chatterbox (23 languages: Arabic, Danish, German, Greek, English, Spanish, Finnish, French, Hebrew, Hindi, Italian, Japanese, Korean, Malay, Dutch, Norwegian, Polish, Portuguese, Russian, Swedish, Swahili, Turkish, Chinese) — the T3 GGUF carries the 2454-token multilingual text tokenizer. Select the language with `-l <code>` (e.g. `-l de`, `-l ja`); English is the default. Two GGUF files are needed: the **T3 model** (text → speech tokens) and the **S3Gen model** (speech tokens → audio). ## Files | File | Quant | Size | Notes | |---|---|---:|---| | `chatterbox-t3-f16.gguf` | F16 | 1.1 GB | T3 AR model — reference quality | | `chatterbox-t3-q8_0.gguf` | Q8_0 | 542 MB | T3 AR model — recommended | | `chatterbox-t3-q4_k.gguf` | Q4_K | 287 MB | T3 AR model — smallest | | `chatterbox-s3gen-f16.gguf` | F16 | 548 MB | S3Gen + vocoder — reference quality | | `chatterbox-s3gen-q8_0.gguf` | Q8_0 | 342 MB | S3Gen + vocoder — recommended | | `chatterbox-s3gen-q4_k.gguf` | Q4_K | 237 MB | S3Gen + vocoder — smallest | Note: vocoder weights (conv_pre, resblocks, conv_post, source fusion) are kept at F32 in all quant levels for audio quality. Quantization applies to the Conformer encoder, UNet decoder, and T3 Llama layers. ## Quick start ```bash # 1. Build CrispASR git clone https://github.com/CrispStrobe/CrispASR cd CrispASR cmake -B build -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=OFF cmake --build build -j --target chatterbox # 2. Pull both model files huggingface-cli download cstr/chatterbox-GGUF chatterbox-t3-q8_0.gguf...