Realtime Transcribe

Captures your microphone, sends chunks to a Whisper-compatible endpoint, and accumulates the result locally. Default chunk length is 10s; default model is qwen3-asr.

Status: idlePending: 0 · Chunks seen: 0

Chunk length

Customs

Model

Default: Qwen3 ASR

silent

Transcript

0 segments

Press Start and speak. The transcript will appear here.

Audio is encoded as 16kHz mono WAV and sent to /api/transcribe. Chunk length is user-controlled (1-600s). Toggle VAD off to send every chunk regardless of voice activity.