Speech to Text
Transcribe audio or video recordings using on-device AI.
The AI model (40–75MB depending on your choice) downloads once on first use and runs entirely on your device's CPU. Transcription takes roughly as long as the recording itself — please keep this tab open until it finishes. Files are limited to 60 minutes and 200MB.
Drop an audio or video file here
or click to browse (MP3, WAV, M4A, MP4, etc.)
Frequently Asked Questions
Is my recording uploaded to a server?expand_more
No. Transcription runs entirely in your browser using OpenAI's Whisper model, compiled to run via WebAssembly. Only the model files themselves are downloaded from a public CDN on first use — your audio or video is never transmitted anywhere.
Which model should I choose?expand_more
Use "Fast" for English-only recordings when you want the quickest result. Use "Accurate" if your recording isn't in English, has background noise, or you simply want the best possible transcription quality.
Can I use the transcript as video subtitles?expand_more
Yes — download the .srt file, which includes timestamps in the standard SubRip format supported by YouTube, video editors, and most media players.