ComfyUI Extension: comfyui-kokoro

Authored by stavsap

Created

Updated

49 stars

Custom ComfyUI Nodes for TTS with Kokoro, genenrate and merge speakers for new style generations.

Custom Nodes (3)

README

Comfy UI Kokoro

<a href="https://www.buymeacoffee.com/stavsapq" target="_blank"><img src="https://cdn.buymeacoffee.com/buttons/default-orange.png" alt="Buy Me A Coffee" height="40" width="174"></a>

<img src="https://img.shields.io/badge/v1.0-green.svg?style=for-the-badge&labelColor=gray&label=Kokoro&color=blue" alt=""/> <img src="https://img.shields.io/badge/0.4.2-green.svg?style=for-the-badge&labelColor=gray&label=Kokoro-onnx&color=blue" alt=""/>

Kokoro TTS nodes, wrapping this kokoro onnx that is based on hexgrad/Kokoro-82M.

workflow.png

note: This picture is also a workflow, just download and drop it into comfy.

Install

Install Via ComfyUI Manager, by stavsap.

img.png

Or

Clone the repo into custom_nodes folder

git clone https://github.com/stavsap/comfyui-kokoro.git

Then cd into comfyui-kokoro, and install requirements.

pip install -r requirements.txt

And finally reboot Comfy.

The onnx model and speakers meta-data will be automatically downloaded on the first run.

If using windows portable version and experience issues with dependencies, check the following:

IMAGE ALT TEXT HERE

Nodes

Currently, there are 3 nodes that can be combined for TTS workflow.

Kokoro Speaker

speaker.png

Select supported speakers.

Kokoro Speaker Combiner

speaker_combiner.png

Combiner node to combine 2 given speakers to new speaker.

  • weight: [1, 0], select the weight of speaker a.

Example:

weight == 0.7 will result in strength of 70% of speaker_a and 30% of speaker_b.

Kokoro Generate

generator.png

  • speaker: input a speaker
  • speed: set the speach speed.
  • lang: set the language, what ever is supported by kokoro.

Available Voices

All supported voices can be found here.

Use Cases:

  1. TTS: Text To Speach, generate voice from test.

  2. Lip Sync: sync lips of videos

lipsync.png

License

  • This repo
  • kokoro-onnx: MIT
  • kokoro model: Apache 2.0

Credits