ComfyUI-Zonos

ComfyUI node to make text to speech audio with your own voices.

** Currently only tested on Windows **

Installation

• You can git clone https://github.com/BuffMcBigHuge/ComfyUI-Zonos into custom_nodes or install from ComfyUI-Manager

cd custom_nodes/ComfyUI-Zonos
git submodule update --init --recursive
pip install -r requirements.txt
git clone https://github.com/Zyphra/Zonos.git

You will need to install eSpeak NG on your machine.

• Windows: You can install eSpeak NG then have to set PHONEMIZER_ESPEAK_LIBRARY=C:\Program Files\eSpeak NG\libespeak-ng.dll in your environment variables, and restart your terminal.
• Linux: You can install via sudo apt install -y espeak-ng.
• You may need to install CUDA Toolkit to use the GPU as well as Visual Studio Build Tools to compile.

How To Use

• Drop a .wav file into ComfyUI/input of a short (5-10s) clear audio of the voice you'd like to use.
• And a .txt file of the same name with what was said.
• Tap "R" in ComfyUI to refresh the node list.
• Use the ZonosGenerate and queue a prompt, (Example Workflow).

Current Issues

• Untested on Mac/Linux
• Model loading isn't handled by Comfy native - ymmv
• Can't get compiling to work - will update when fixed

Special Thanks

• Zyphra for the Zonos model.
• niknah for the F5-TTS node.
• sdbds for the Zonos-for-windows gradio_interface.py