have your chat

in your earphones

TikTok Live chat text-to-speech in your browser
News
60M TOKENS60 million tokens or (20k pages written text) later the Vercel Fishaudio promo endedMUGGS IS ALIVEsurstreaming merch is here, starting with mugs and tees — printed on demand near you and shipped worldwide by GelatoSPECIAL VOICEGive your special chatters an own Voice! Assign a custom voice to any user in your chat.TWO-PERSON STREAMThe stream needs eyes. Chat doesn't — let someone hold the screen while you read your whole live chat in your ears.THREAD AHEADFinally work with Multithreaded Supertonic3 engine, making everything responsive while number crunching!S2.1 PRO FREEGood news from Fish Audio: s2.1-pro-free stays free — the same model, the same 83 languages, $0. It is the default in your Dashboard now, so the engine just works.HOW TO USE FISH AUDIO Switch the TTS Engine in your Dashboard to try it. Paste your own Fish Audio API key in the Fish menu to read chat, and load the voices you trained in there too.

Live TTS from livestream chat

Comments and gifts can be read aloud with filters, blocklist and per-user voices. Connect your TikTok, Twitch and Kick livestreams

31 languages with 10 voices

High-fidelity Speech that's generated entirely on your own local device hardware. Optionally use Fish Audio with 80 languages and Voicecloning features

No latency

Supertonic3 runs on WebAssembly (Wasm) & WebGPU to start delivering audio in under 200ms from the chat into your headphones

Stream events data export

Download the structured JSON file of all your stream events, comments and gifts to save it forever

We log nothing

The service logs nothing. Chat stays in your browser, is used to synthesize speech and is discarded right after — nothing is stored, nothing is written to logs

Under the hood

Built on a lean, well-known stack:

  • Supertonic 3 — ONNX speech model on Wasm + WebGPU, in your browser
  • Next.js — React framework handles event processing and responsive UI
  • Litestar — Python lifts Auth and SSE Pump to the browser
  • PostgreSQL for persistent data
  • Redis for queues, caches and rate limiting

The stream needs eyes. Chat doesn't.

Split a live in two: someone holds the picture, someone holds the room. Built for blind and low-vision streamers, and for anyone with a co-host or a second monitor.

How the two-person stream works →
With your consent we measure how the site is used: your browser stores a random visitor ID and sends page views to our own analytics server in Frankfurt, which records your IP address, browser and approximate location. Disabling keeps all analytics off. You can change your choice anytime under “Cookie settings” in the footer. Privacy Policy