---
title: DialogueSidon
canonical_url: "https://www.modelscope.cn/models/mlx-community/DialogueSidon"
md_url: "https://www.modelscope.cn/models/mlx-community/DialogueSidon.md"
repository: mlx-community/DialogueSidon
last_updated: 2026-09-06
license: cc-by-nc-4.0
model_type:
  - dialogue_sidon
base_model:
  - sarulab-speech/DialogueSidon
base_model_relation: finetune
parameters: 444.2M
tensor_type:
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 4
stars: 0
tags:
  - mlx
  - audio
  - audio-source-separation
---

# DialogueSidon

> DialogueSidon - mlx-community 在 ModelScope 开源的模型。Two-speaker dialogue separation and restoration, with 24 kHz mono stems. Converted from sarulab-speech/DialogueSidon for use with mlx-audio.

mlx-community/DialogueSidon 是 ModelScope 魔搭社区上的 444.2M 参数机器学习模型，采用 cc-by-nc-4.0 许可，基于 sarulab-speech/DialogueSidon 构建。

- **Repository**: mlx-community/DialogueSidon
- **License**: cc-by-nc-4.0
- **Parameters**: 444.2M
- **Base model**: sarulab-speech/DialogueSidon
- **Tags**: mlx, audio, audio-source-separation
- **Downloads**: 4
- **Stars**: 0
- **Last updated**: 2026-09-06

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

---

# DialogueSidon

Two-speaker dialogue separation and restoration, with 24 kHz mono stems.
Converted from [sarulab-speech/DialogueSidon](https://huggingface.co/sarulab-speech/DialogueSidon/tree/d43d7478402a5527136c6733c3f4359c37b312ab) for use with [mlx-audio](https://github.com/Blaizzy/mlx-audio).

```python
from mlx_audio.sts import load
from mlx_audio.audio_io import write

model = load("mlx-community/DialogueSidon")
result = model.separate("dialogue.wav", num_steps=30, seed=0)
for i, speaker in enumerate(result.speakers, 1):
    write(f"speaker_{i}.wav", speaker, result.sample_rate)
```

The two channels are anonymous speaker slots. Long recordings use overlapping
chunks and waveform correlation to maintain channel assignment. Consistency
across silence is not guaranteed. This is an offline generative model, so the
stems need not sum to the mixture. The default chunk and overlap are 20 and 5
seconds; use `chunk_seconds=None` for whole-file inference when memory permits.

The checkpoint retains **CC-BY-NC-4.0**. Original model by Wataru Nakata,
Yuki Saito, Kazuki Yamauchi, Emiru Tsunoo, and Hiroshi Saruwatari (SaruLab).
See the [original model card](https://huggingface.co/sarulab-speech/DialogueSidon) and
[Sidon implementation](https://github.com/sarulab-speech/Sidon).