---
license: apache-2.0
language:
- zh
- en
- ja
- ko
- de
- fr
- es
- id
- it
- th
- pt
- ru
- ms
- vi
base_model: netease-youdao/Confucius4-TTS
pipeline_tag: text-to-speech
library_name: mlx
tags:
- mlx
- apple-silicon
- text-to-speech
- voice-cloning
---

# Confucius4-TTS-mlx

Full-precision (fp32) MLX build of [netease-youdao/Confucius4-TTS](https://huggingface.co/netease-youdao/Confucius4-TTS)
(multilingual + cross-lingual zero-shot voice cloning, 14 languages: zh, en, ja,
ko, de, fr, es, id, it, th, pt, ru, ms, vi) for Apple Silicon.

All weights kept in fp32 (T2S, w2v-bert encoder, S2A flow, BigVGAN vocoder).
~5.4 GB total. For a smaller/faster 8-bit build see
[mlx-community/Confucius4-TTS-mlx-int8](https://huggingface.co/mlx-community/Confucius4-TTS-mlx-int8).

- **T2S:** ~2.5 GB
- **w2v-bert:** ~1.5 GB
- **S2A + BigVGAN:** ~0.8 GB
- **Speed (Apple M5):** RTF ~2.4

## Usage

Needs the `confucius4` model in [mlx-audio](https://github.com/Blaizzy/mlx-audio)
(PR [#799](https://github.com/Blaizzy/mlx-audio/pull/799)):

```python
from mlx_audio.tts.utils import load
model = load("mlx-community/Confucius4-TTS-mlx")
for r in model.generate("Xin chào", ref_audio="voice.wav", lang="vi"):
    ...  # r.audio at 22050 Hz
```

## Attribution & license
- Model & architecture: netease-youdao/Confucius4-TTS (Apache-2.0)
- Vocoder: NVIDIA BigVGAN v2; speaker encoder: 3D-Speaker CAMPPlus (funasr)
- MLX port by [Hert4](https://github.com/Hert4), released under Apache-2.0.