Live interpreter

Live captions, translation, meeting notes and a spoken interpretation, in one page. Listen to a meeting tab or speak yourself: recognition (Qwen3-ASR), translation (Qwen3 4B), notes (DeepSeek) and voice (Qwen3-TTS) all run on Babel Mesh nodes.

Mode

Translate into Audio from

Speak the translation aloud Correct recognition errors (context and web search) Keep recordings in this browser to recognise again later Save the session to my Hive library (needs e-mail sign-in and the sync passphrase) Write meeting notes every 45 seconds Voice

Pick a mode and press Start. For a meeting, choose the tab and tick “Share tab audio”.

Captions

Notes

Notes appear here once there is something to summarise.

Earlier sessions in this browser

Audio goes to the speech engine while you listen; text goes to the translation and notes models. Nothing is stored on the server. Sessions are kept only in this browser (the last 10), and you can delete them. A browser cannot create a virtual microphone, so the spoken translation plays on your speakers: use a headset in “I speak” mode so it does not feed back. Recognition handles Chinese and English speech; other languages are translated from what was recognised. Sharing a tab’s sound works in Chrome and Edge.