Sign in

Nodes Browser

ComfyDeploy: How ComfyUI_StepAudioTTS works in ComfyUI?

What is ComfyUI_StepAudioTTS?

A Text To Speech node using Step-Audio-TTS in ComfyUI. Can speak, rap, sing, or clone voice.

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_StepAudioTTS and select it
  5. Close the build step dialig and then click on the "Save" button to rebuild the machine

中文 | English

A Text To Speech node using Step-Audio-TTS in ComfyUI. Can speak, rap, sing, or clone voice.

Update

[2025-03-21] ⚒️: Completely refactored the code, added more tunable parameters, and max_length can be adjusted according to the text length. Optional unload_model to choose whether to unload the model to accelerate inference speed.

[2025-03-07]⚒️: Custom speakers can be defined directly in ComfyUI\models\TTS\Step-Audio-speakers\speakers_info.json without the need for input in the node.

Move the Step-Audio-speakers folder from this repository to the ComfyUI\models\TTS folder. The structure is as follows:

ComfyUI\models\TTS
├── Step-Audio-Tokenizer
├── Step-Audio-speakers
├── Step-Audio-TTS-3B

You can then freely customize speakers under the ComfyUI\models\TTS\Step-Audio-speakers folder for use. Ensure that the speaker name configuration matches exactly:

[2025-03-06]⚒️: New recording node MW Audio Recorder can be used to record audio with a microphone, and the progress bar displays the recording progress:

| 参数名/Parameter | 作用描述/Description | 范围/Range | 注意事項/Notes | |---------------------|------------------------------------------------------------------------------------|--------------------------------|------------------------------------------------------------------------------| | trigger | 录音触发开关 - 设为True开始录音<br>Recording trigger - Set to True to start recording | Boolean (True/False) | 需要从False切到True才能触发<br>Requires changing from False to True to activate | | record_sec | 主录音时长（秒）<br>Main recording duration (seconds) | 1-60 (整数/integer) | 实际时长<br>Actual duration | | n_fft | FFT窗口大小（影响频率分辨率）<br>FFT window size (affects frequency resolution) | 512,1024,...,4096 (512倍数/multiplies) | 值越大频率分辨率越高<br>Higher values give better frequency resolution | | sensitivity | 降噪灵敏度（值越高越激进）<br>Noise reduction sensitivity (higher=more aggressive) | 0.5-3.0 (步长0.1/step 0.1) | 1.2=标准办公室环境<br>1.2=standard office environment | | smooth | 时频平滑系数（值越高越自然）<br>Time-frequency smoothing (higher=more natural) | 1,3,5,7,9,11 (奇数/odd numbers) | 建议语音：5，音乐：7<br>Recommended: 5 for speech, 7 for music | | sample_rate | 采样率（影响音质与文件大小）<br>Sampling rate (affects quality & size) | 16000/44100/48000 Hz | 44100=CD音质<br>44100=CD quality |

[2025-03-02]⚒️: Add experimental custom_mark, surrounding with "()", for example (温柔)(东北话), it may have an effect.

[2025-02-25]⚒️: Support custom speaker custom_stpeaker.

Installation

cd ComfyUI/custom_nodes
git clone https://github.com/billwuhao/ComfyUI_StepAudioTTS.git
cd ComfyUI_StepAudioTTS
pip install -r requirements.txt

# python_embeded
./python_embeded/python.exe -m pip install -r requirements.txt

Model Download

Download to the ComfyUI\models\TTS folder

Huggingface

| Models | Links | |-------|-------| | Step-Audio-Tokenizer | 🤗huggingface | | Step-Audio-TTS-3B | 🤗huggingface |

Modelscope

| Models | Links | |-------|-------| | Step-Audio-Tokenizer | modelscope | | Step-Audio-TTS-3B | modelscope |

Supports Chinese, English, Korean, Japanese, Sichuanese, Cantonese etc.

Acknowledgements

Part of the code for this project comes from:

Thank you to all the open-source projects for their contributions to this project!