Translator

Demo limit: 60 seconds per recording so the in-browser speech model stays responsive. Stop anytime; recording auto-stops at the limit.

Using the slower WASM speech model — WebGPU isn't available in this browser.

Demo limits: up to 60 seconds and 50 MB. We extract the audio in your browser, transcribe it locally, then translate. Longer or larger files would freeze the tab on most laptops while the local model warms up.

Strict mode: tokens whose sign isn't reviewed by a Deaf native signer of the target language are spelled out letter-by-letter instead of rendered from the unreviewed sign. Off by default. Your preference is remembered in this browser.