ComfyUI_FL-ClearVoice

ComfyUI_FL-ClearVoice
★ 18

语音增强超分辨率去噪恢复模型自动下载
为ComfyUI提供音频增强节点，支持语音增强、超分、去噪与修复，兼容多后端并自动下载模型。
💡 在ComfyUI流程中一键提升语音和音频质量。
🍴 2 Forks💻 Python🔄 2026-01-13
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/e00a65475347
📦 requirements.txt
#
FL
ClearVoice
Dependencies
#
===========================
#
Core
dependencies
clearvoice
torch>=2.0.0
torchaudio>=2.0.0
numpy>=1.24.0
soundfile>=0.12.0
librosa>=0.10.0
scipy>=1.10.0
#
Pin
transformers
to
avoid
PyTorch
2.1.x
compatibility
issues
#
(newer
versions
require
PyTorch
2.4+)
transformers<4.46.0
#
Optional:
Resemble-Enhance
(for
Resemble_Enhance
and
Resemble_Denoise
models)
#
Install
with:
pip
install
resemble-enhance
#
resemble-enhance
#
Optional:
VoiceFixer
(for
VoiceFixer
model)
#
Install
with:
pip
install
voicefixer
#
voicefixer
Workflow Preview
📄 README

FL ClearVoice

Audio enhancement nodes for ComfyUI featuring speech enhancement, super-resolution, denoising, and restoration using multiple AI backends.

[](https://github.com/modelscope/ClearerVoice-Studio)

[](https://www.patreon.com/Machinedelusions)

Features

  • Speech Enhancement – Remove noise and improve speech clarity
  • Super-Resolution – Upscale low-quality audio to higher sample rates
  • Audio Restoration – Fix degraded or compressed audio
  • Multiple Backends – ClearVoice, Resemble-Enhance, and VoiceFixer models
  • Automatic Resampling – Handles sample rate conversion automatically
  • Mono/Stereo Support – Preserves original channel configuration
  • Centralized Model Storage – All models download to ComfyUI/models/clear_voice/
  • Real-time Download Progress – Progress bars with speed and ETA for model downloads

    • Nodes

    | Node | Description |

    |——|————-|

    | Model Loader | Downloads and caches enhancement models |

    | Process | Processes audio using the loaded model |

    Installation

    ComfyUI Manager

    Search for “FL ClearVoice” and install.

    Manual

    cd ComfyUI/custom_nodes
git clone https://github.com/filliptm/ComfyUI_FL-ClearVoice.git
cd ComfyUI_FL-ClearVoice
pip install -r requirements.txt

    Quick Start

  • Add FL ClearVoice Model Loader and select a model
  • Connect to FL ClearVoice Process node
  • Provide input audio
  • Run the workflow

    • Models

    ClearVoice Models

    | Model | Type | Output SR | Notes |

    |——-|——|———–|——-|

    | MossFormer2_SE_48K | Enhancement | 48kHz | High quality speech enhancement |

    | FRCRN_SE_16K | Enhancement | 16kHz | Fast speech enhancement |

    | MossFormerGAN_SE_16K | Enhancement | 16kHz | GAN-based enhancement |

    | MossFormer2_SR_48K | Super-Resolution | 48kHz | Audio upscaling |

    Resemble-Enhance Models

    | Model | Type | Output SR | Notes |

    |——-|——|———–|——-|

    | Resemble_Enhance | Full Enhancement | 44.1kHz | Denoise + enhance + vocoder |

    | Resemble_Denoise | Denoise Only | 44.1kHz | Fast denoising |

    VoiceFixer

    | Model | Type | Output SR | Notes |

    |——-|——|———–|——-|

    | VoiceFixer | Restoration | 44.1kHz | All-in-one audio restoration |

    Models download automatically on first use to ComfyUI/models/clear_voice/:

    ComfyUI/models/clear_voice/
├── clearvoice/           # ClearVoice models
│   ├── MossFormer2_SE_48K/
│   ├── FRCRN_SE_16K/
│   └── ...
├── resemble_enhance/     # Resemble-Enhance models
└── voicefixer/           # VoiceFixer models

    Requirements

  • Python 3.10+
  • 8GB RAM minimum (16GB+ recommended)
  • NVIDIA GPU recommended (CPU and Mac supported)

    • Platform Notes

  • NVIDIA GPU: Full acceleration for all models
  • Mac (MPS): ClearVoice uses MPS, Resemble-Enhance falls back to CPU due to conv1d channel limits
  • CPU: All models supported, slower processing

    • License

    MIT