Chatterbox Nano ONNX
KitsuMate mirror of the independent
owensong/chatterbox-nano-ONNX
conversion for reproducible Unity and ONNX Runtime downloads. This is not an
official Resemble AI ONNX release.
Contents
The repository contains the conversion's only complete published profile:
| Stage | File | Precision |
|---|---|---|
| Text and speech-token embeddings | embed_tokens_fp16.onnx |
FP16 |
| Reference speech encoder | speech_encoder_q4f16.onnx |
Q4F16 |
| Autoregressive language model | language_model_q4f16.onnx |
Q4F16 |
| Conditional acoustic decoder | conditional_decoder_q4.onnx |
Q4 |
Each graph must remain beside its matching .onnx_data file. This mixed profile
is identified as q4 by KitsuMate because the Q4 decoder is its largest
component.
The conversion has passed an end-to-end CPU smoke test, but has not received a representative speech-quality or device-compatibility evaluation. It is non-streaming and supports English only.
Reproducibility
- Mirrored ONNX conversion revision:
4a66d7dab72a9e98f24b515d49a1d7a81632df2e - Official Nano checkpoint revision:
71ccd1d0081b430592cea481f4307e764e07bc64 - Official implementation:
resemble-ai/chatterbox - Official model:
ResembleAI/chatterbox-nano - Independent ONNX conversion and packaging:
owensong - File hashes and byte sizes:
SHA256SUMS
default_voice.wav is redistributed unchanged from
onnx-community/chatterbox-ONNX
revision 3cab09af388d3f02bba43443fce88c1f4525ac43, under the same MIT license. Its
SHA-256 is 3ebc531cdaba358a327099c1c4f0448026719957bcf4d8e9868767f227e02f4e.
License and attribution
Chatterbox and its original model weights are Copyright (c) 2025 Resemble AI
and licensed under the MIT License. The ONNX conversion and packaging work is
attributed to owensong. This mirror preserves both upstream notices.
- Downloads last month
- 50