Sign in

Nodes Browser

ComfyDeploy: How ComfyUI-FunAudioLLM works in ComfyUI?

What is ComfyUI-FunAudioLLM?

Comfyui custom node for [a/FunAudioLLM](https://funaudiollm.github.io/) include [a/CosyVoice](https://github.com/FunAudioLLM/CosyVoice) and [a/SenseVoice](https://github.com/FunAudioLLM/SenseVoice).

How to install it in ComfyDeploy?

Head over to the machine page

  1. Click on the "Create a new machine" button
  2. Select the Edit build steps
  3. Add a new step -> Custom Node
  4. Search for ComfyUI-FunAudioLLM and select it
  5. Close the build step dialig and then click on the "Save" button to rebuild the machine
<!-- * @Author: SpenserCai * @Date: 2024-10-04 11:28:28 * @version: * @LastEditors: SpenserCai * @LastEditTime: 2024-10-06 11:56:47 * @Description: file content -->

ComfyUI-FunAudioLLM

Comfyui custom node for FunAudioLLM include CosyVoice and SenseVoice

Features

CosyVoice

  • CosyVoice Version: 2024-10-04
  • Support SFT,Zero-shot,Cross-lingual,Instruct
  • Support CosyVoice-300M-25Hz in zero-shot and cross-lingual
  • Support SFT's 25Hz(unoffical)
  • <details> <summary>Save and load speaker model in zero-shot</summary> <img src="./assets/SaveSpeakerModel.png" alt="zh-CN" /> <br> <img src="./assets/LoadSpeakerModel.png" alt="zh-CN" /> </details>

SenseVoice

  • SenseVoice Version: 2024-10-04
  • Support SenseVoice-Small
  • <details> <summary>Support Punctuation segment (need turn off use_fast_mode)</summary> <img src="./assets/SenseVoice.png" alt="zh-CN" /> <br> <img src="./assets/PuncSegment.png" alt="zh-CN" /> </details>

How use

apt update
apt install ffmpeg

## in ComfyUI/custom_nodes
git clone https://github.com/SpenserCai/ComfyUI-FunAudioLLM
cd ComfyUI-FunAudioLLM
pip install -r requirements.txt

Windows

In windows need use conda to install pynini

conda install -c conda-forge pynini=2.1.6
pip install -r requirements.txt

MacOS

If meet error when you install

brew install openfst
export CPPFLAGS="-I/opt/homebrew/include"
export LDFLAGS="-L/opt/homebrew/lib"
pip install -r requirements.txt

If your network is unstable, you can pre-download the model from the following sources and place it in the appropriate directory.

WorkFlow

<img src="./assets/Workflow_FunAudioLLM.png" alt="zh-CN" />