Comfy UI Kokoro

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

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

Install

Install Via ComfyUI Manager, by stavsap.

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:

Nodes

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

Kokoro Speaker

Select supported speakers.

Kokoro Speaker Combiner

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

• 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

License

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

Credits

• Kokoro TTS Engine
• ComfyUI
• ComfyUI-BS_Kokoro-onnx
• ComfyUI-KokoroTTS