Text to Speech
Powered by Kokoro-82M by hexgrad, built on StyleTTS 2. Running locally in your browser via kokoro-js and Transformers.js. Your data never leaves your browser.
The quality is dramatically worse than large server-side models, but considering it runs on-device, unlimited and free, it’s still quite impressive. In-browser models will likely improve significantly with WebGPU support.
Ready. Model will download (~92 MB) when you click Generate.