Drop an audio file in, or click to browse. It stays on this device — transcription runs locally, nothing is sent anywhere.
Add up to ten photos and/or video clips to play behind the captions, in sequence. Photos get a slow, randomized Ken Burns pan & zoom; videos play as-is. Drag inside each frame to reposition, and use the slider to zoom. If you skip this, the tool falls back to a plain background.
Runs OpenAI's Whisper model right in this tab (first run downloads and caches the model — a one-time wait). Audio is auto-normalized and trimmed of dead silence before transcribing. Review and correct the text below before rendering — a bigger model, beam search, and the dictionary below all help, but a human pass still catches the rest.
Whisper will mishear names and jargon consistently — fix them once here instead of every transcript. One per line: misheard text => correct text.
Captions are drawn in sync with the audio's own clock, then the whole thing is recorded into a downloadable video file.
Seconds before the narration and captions kick in — the background keeps playing during that lead-in. Leave at 0 to start immediately.
Or drag the gold box on the preview to place captions exactly where you want them.
Recording format depends on what your browser supports — this tool prefers MP4 and falls back to WebM automatically. WebM plays everywhere in-browser and on most platforms; convert to MP4 afterward if a specific tool requires it.