---
title: supertonic-2
canonical_url: "https://www.modelscope.cn/models/mlx-community/supertonic-2"
md_url: "https://www.modelscope.cn/models/mlx-community/supertonic-2.md"
repository: mlx-community/supertonic-2
last_updated: 2026-04-05
model_type:
  - supertonic
parameters: 61.4M
tensor_type:
  - F32
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 8
stars: 0
---

# supertonic-2

> supertonic-2 - mlx-community 在 ModelScope 开源的模型。mlx-community/supertonic-2

- **Repository**: mlx-community/supertonic-2
- **Parameters**: 61.4M
- **Downloads**: 8
- **Stars**: 0
- **Last updated**: 2026-04-05

Source: https://www.modelscope.cn/models/mlx-community/supertonic-2

---

# mlx-community/supertonic-2

This model was converted to MLX format from [`Supertone/supertonic-2`](https://huggingface.co/Supertone/supertonic-2) using mlx-audio version **0.2.8**.

SuperTonic 2 is a high-quality text-to-speech model with voice style control.

## Use with mlx-audio

```bash
pip install -U mlx-audio
```

### CLI Example:
```bash
mlx_audio.tts.generate --model mlx-community/supertonic-2 --text "Hello, this is a test." --voice M1
```

### Python Example:
```python
from mlx_audio.tts.utils import load_model

model = load_model("mlx-community/supertonic-2")
for result in model.generate("Hello, this is a test.", voice="M1"):
    print(f"Generated {result.audio_duration} of audio")
```

## Model Details

- **Architecture**: Text encoder + Duration predictor + Flow matching (vector field) + Vocoder
- **Sample rate**: 44100 Hz
- **Voices**: M1-M5, F1-F5 (10 built-in voice styles)
- **Latent dim**: 24 (compressed to 144 via chunking)
- **Flow matching steps**: 10 (configurable)