News2025August 2025

Kokoro TTS runs in the browser via WebGPU

Audio2025-08-28 Source: github.com

A tiny, high-quality text-to-speech model runs client-side with no server, through WebGPU.

Speech synthesis has shrunk to the point where a capable voice runs entirely in a browser tab over WebGPU — no server, no install — good for accessibility and offline tools.

See how it's done on the WebGPU page and the rest of the models on audio.

Read the original at github.com ↗ More from August 2025