Capture audio

The app loads the model without starting capture. Choose a source only when you are ready.

Microphone

Select Microphone, then Start microphone. Approve the macOS microphone prompt. Stop and finish drains queued audio and commits the final model token before ending capture.

Browser or YouTube tab

Select YouTube / tab, then choose a browser tab and enable its audio in the share picker. WebKit and Chrome present different share pickers; macOS Screen & System Audio Recording permission may be required. If the native WebKit picker does not offer tab audio, use the source launcher described in Development and open the loopback page in Chrome.

Audio or video file

Select Audio / video, choose a local file, then Transcribe file. FFmpeg decodes the file locally to mono 16 kHz audio. Temporary uploads are deleted after completion, cancellation, disconnect, or error.

Only one live or file transcription runs at a time. Live output is append-only; Copy, Export .txt, and Pop out captions act on the committed text.