Kokoro TextToSpeech Node for ComfyUI

A custom node for ComfyUI that provides Text-to-Speech capabilities using the Kokoro TTS engine.

The basic TTS

TTS with LatentSync for Lipsync

https://github.com/user-attachments/assets/55a1cd18-ec8f-4127-8ee6-62c68e493b30

Example Result.

Features

• High-quality text-to-speech synthesis
• Multiple voice options
• Support for multilingual text
• Easy integration with ComfyUI workflows

Installation

1. Clone this repository into your ComfyUI custom nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/benjiyaya/ComfyUI-KokoroTTS

2. Download required model files:

   • Create a folder Kokorotts under ComfyUI/models
   • Go to https://huggingface.co/thewh1teagle/Kokoro/tree/main
   • Download the model 'kokoro-v0_19.onnx' file and save to 'Kokorotts' folder
   • Download the voices 'voices.json' file and save to 'Kokorotts' folder
   • Place both files in the ComfyUI/models/Kokorotts directory

3. Install required Python packages:

pip install -r requirements.txt

or

if you are using window protable version.

Go to 'ComfyUI_windows_portable' folder
 run the command : "python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\ComfyUI-KokoroTTS\requirements.txt"

Available Voices

The following voices are available:

• af (African Female)
• af_sarah (African Female Sarah)
• af_bella (African Female Bella)
• af_nicole (African Female Nicole)
• af_sky (African Female Sky)
• am_adam (African Male Adam)
• am_michael (African Male Michael)
• bf_emma (British Female Emma)
• bf_isabella (British Female Isabella)
• bm_george (British Male George)
• bm_lewis (British Male Lewis)

Usage

1. In ComfyUI, locate the "Kokoro TextToSpeech" node under the "kokoro" category
2. Connect the node to your workflow
3. Input your text and select a voice
4. The node will output an audio waveform that can be used with other audio nodes

Input Parameters

• text: The text you want to convert to speech (supports multiline text)
• speaker: The voice to use for speech synthesis (default: af_sarah)

Output

• audio: Audio data in the format expected by ComfyUI audio nodes

Error Handling

The node includes comprehensive error handling for common issues:

• Missing model or voice files
• Invalid text input
• TTS generation failures

Error messages will be logged with detailed information to help troubleshoot any issues.

License

kokoro-onnx: MIT kokoro model: Apache 2.0

Credits

• Kokoro TTS Engine: [Include credits for the original Kokoro TTS project]
• ComfyUI: https://github.com/comfyanonymous/ComfyUI
• ComfyUI-BS_Kokoro-onnx https://github.com/Burgstall-labs/ComfyUI-BS_Kokoro-onnx