cstr/lahgtna-chatterbox-v1-GGUF

lahgtna-chatterbox-v1 — GGUF (ggml-quantised)

GGUF / ggml conversion of oddadmix/lahgtna-chatterbox-v1 for use with CrispStrobe/CrispASR.

Arabic T3 (text-to-speech-tokens) variant of Chatterbox TTS. The T3 language model is finetuned for Arabic text while sharing the same S3Gen vocoder as the base English model. Distributed under MIT license.

This repo contains only the T3 model (Arabic text → speech tokens). You also need the shared S3Gen model from cstr/chatterbox-GGUF to produce audio.

Files

File	Quant	Size	Notes
chatterbox-t3-f16.gguf	F16	1.1 GB	Arabic T3 — reference quality

For S3Gen (vocoder), download from the base Chatterbox repo:

huggingface-cli download cstr/chatterbox-GGUF chatterbox-s3gen-q8_0.gguf --local-dir .

Architecture

Same as base Chatterbox but with Arabic-trained T3:

• T3: 30-layer Llama AR (1024D, 16 heads, RoPE, SwiGLU, CFG)
• Multilingual tokenizer (704 text tokens + Arabic graphemes)
• S3Gen + HiFTGenerator vocoder shared with base English model

Related models

• cstr/chatterbox-GGUF — Base English Chatterbox (required for S3Gen)
• oddadmix/lahgtna-chatterbox-v1 — Original weights

Conversion

python models/convert-chatterbox-to-gguf.py \
    --input oddadmix/lahgtna-chatterbox-v1 \
    --output-dir . --t3-only

License

MIT — same as upstream oddadmix/lahgtna-chatterbox-v1.

Provenance and EU AI Act Art. 53 note

• Upstream model: oddadmix/lahgtna-chatterbox-v1 — published by oddadmix.
• Upstream licence: mit. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (GGUF/GGML). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository. No training-content summary was found on the upstream model card at the time of writing; that documentation gap is upstream's and is not filled here.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.