Turn speech into an editable transcript and subtitles — SRT, VTT, or burned into the video. Nothing is uploaded: the speech model runs inside this tab.
Drop an audio or video file hereor tap to choose — MP3, M4A, WAV, MP4, MOV, WebMChoose a file
Transcribe
Honestly: Whisper lists Luxembourgish, but at these sizes it gets about nine words in ten wrong — measured on FLEURS recordings, both models. What comes out reads like German guesses. It can show where someone speaks, not what they said.
Transcript
Tap a time to jump there. Edit any line; an emptied line is left out of the files.
Burn into video
Draws the lines onto the frames and records the video again as it plays — so a three-minute video takes three minutes. Keep this tab in front while it runs.
The file never leaves this tab. The speech model — Whisper, by OpenAI — is downloaded once from this site and runs on your own device; the sound is not sent anywhere, and works offline after the first visit.