Repolode
Submit
Saganaki22

Saganaki22/ComfyUI-Ming-3-tts

A native ComfyUI custom node that brings the Ming-3.0-tts speech generation models (0.5B and Tiny 9B) into ComfyUI workflows.

View on GitHubWebsite
Stars
11
Forks
1
Open issues
0
Last commit
yesterday

What is it?

ComfyUI-Ming-3-tts is a native ComfyUI port of the Ming-3.0-tts text-to-speech models, covering the 0.5B and Tiny 9B variants. It adds nodes for loading a model, generating speech, building multi-speaker podcasts with the Tiny model, and transcribing reference audio with local Whisper, so speech generation can be composed with standard ComfyUI audio preview and save nodes.

The generation node supports several tasks selected per run: voice cloning from a reference recording and its transcript, voice design from a text description, named built-in voices, freeform delivery instructions, basic speed/pitch/volume control, emotion, and Cantonese or Sichuanese dialect. Inline vocal tags such as [Laughter] and [Sigh] can be placed in the text. Outputs are mono 44.1 kHz audio.

Models are downloaded from a Hugging Face repository in original or INT8 ConvRot form, or loaded from existing local folders and extra model paths. The project includes example workflows covering each task, and supports CUDA, with CPU execution available.

Key features

  • 01Loads Ming-3.0-tts 0.5B or Tiny 9B checkpoints, original or INT8 ConvRot.
  • 02Clones a voice from a reference recording and its transcript.
  • 03Supports built-in voices, voice design, emotion, dialect and freeform delivery instructions.
  • 04Builds two or more speaker podcasts with the Tiny 9B model.
  • 05Transcribes reference audio locally with Whisper.
  • 06Includes example workflows for each task.

Use cases

  • Adding text-to-speech narration or voiceovers to an existing ComfyUI image or video pipeline.
  • Cloning a specific voice from a short recording for generated speech.
  • Producing multi-speaker dialogue or podcast-style audio from a script.
  • Experimenting with Chinese dialect or emotional speech generation in a node-based workflow.

Topics

#text-to-speech#voice-cloning#comfyui#audio#speech-synthesis#whisper#ming#ming-tts#one-shot#quantization

FAQ

What is ComfyUI-Ming-3-tts?
A native ComfyUI custom node that brings the Ming-3.0-tts speech generation models (0.5B and Tiny 9B) into ComfyUI workflows. ComfyUI-Ming-3-tts is a native ComfyUI port of the Ming-3.0-tts text-to-speech models, covering the 0.5B and Tiny 9B variants. It adds nodes for loading a model, generating speech, building multi-speaker podcasts with the Tiny model, and transcribing reference audio with local Whisper, so speech generation can be composed with standard ComfyUI audio preview and save nodes.
Is ComfyUI-Ming-3-tts free to use? What license does it have?
ComfyUI-Ming-3-tts is open source under the MIT license. Check the license terms before using it in commercial products.
Is ComfyUI-Ming-3-tts still maintained?
Actively maintained. The last commit to ComfyUI-Ming-3-tts was on Oct 20, 2026.
What can I use ComfyUI-Ming-3-tts for?
Adding text-to-speech narration or voiceovers to an existing ComfyUI image or video pipeline. Cloning a specific voice from a short recording for generated speech. Producing multi-speaker dialogue or podcast-style audio from a script. Experimenting with Chinese dialect or emotional speech generation in a node-based workflow.
Where can I get ComfyUI-Ming-3-tts?
The source code and installation instructions are on GitHub: https://github.com/Saganaki22/ComfyUI-Ming-3-tts

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects