Live speech-to-text
Qwen3-ASR running on a Babel Mesh node. Speak for up to 45 seconds, or upload a recording. The text appears as the audio comes in, and nothing is stored.
Ready. Chinese and English work; the browser sends 16 kHz mono audio.
Audio is sent to the speech engine for one request and is not stored or logged. The demo takes one recording at a time, up to 45 seconds. Results depend on how busy the node is.