Skip to content

kokoro-onnx

A lightweight ONNX TTS project with English voices

Visit official site

What kokoro-onnx is

The project is described as a text-to-speech system built on the Kokoro model and the ONNX runtime. It supports English, offers several voices including whisper, and stays lightweight at about 300 MB.

What you can do with it

  • Generate speech from English text
  • Pick from multiple voices including whisper
  • Run the model through the ONNX runtime for deployment
  • Start quickly with the provided example scripts

Application scenarios

  • Add voice prompts to a mobile app
  • Integrate natural speech into assistant devices
  • Explore new speech generation techniques in research

Getting started

  • Install uv or use a regular Python environment
  • Create a project folder and run uv init -p 3.12
  • Add kokoro-onnx and soundfile with uv add
  • Copy examples/save.py into hello.py

Pros & cons

✓ What we like

  • MIT license and ONNX runtime ease integration
  • Example scripts help users get started

! What to watch out for

  • Named language support is English, with others planned
  • Setup is documented with uv and Python 3.12

FAQ

How large is the model?

The material states about 300 MB, or about 80 MB after quantization.

What license and languages apply?

The project uses the MIT license and lists English, with French, Japanese, Korean and Chinese planned.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review