Turns speech in a recording into text, with a timestamp on every phrase: interviews, lectures, meetings, voice notes, podcasts, or the sound of a video. It runs Whisper - OpenAI’s speech recognition model, the one most transcription services are built on - inside this page, so the recording is never uploaded. The transcript can be read alongside the audio, with each timestamp jumping to that moment, and corrected on the page, then saved as plain text, as a Word document, or as SRT or VTT subtitles for a video.
Whisper recognises 99 languages and can detect which one it is hearing, or translate what it hears into English. Speed depends on your computer. In testing, the Balanced model worked through a minute of speech in about fifteen seconds on the processor alone, and faster where the browser could use the graphics card; a slow laptop will take longer.
- Is my recording uploaded to be transcribed?
- No. The audio is decoded by your browser and the model runs on your own machine. The only download is the model itself, from huggingface.co, the first time - that tells them which model size you chose and nothing about what you are transcribing.
- Which model should I choose?
- Fast is the smallest and quickest, and fine for clear speech. Balanced is noticeably more accurate for a modest extra download, and is the default. Accurate handles accents, noise and specialist words best, but it is about 250 MB and slow without a graphics card.
- How do I make subtitles for a video?
- Open the video file itself - the sound is taken from it - transcribe, and press .srt. Almost every video editor and player, and YouTube, accepts an SRT file. Each subtitle is one phrase with the times Whisper gave it, so check a few before publishing.
- Can I transcribe a voice note or voice message?
- Yes. iPhone Voice Memos (.m4a), Android recorder files and voice messages saved from WhatsApp or Telegram (.opus or .ogg) all open here, as does any recording your browser can play. Save the voice message from the chat to your phone or computer first, then choose it.
- Does it tell different speakers apart?
- No. Whisper writes what was said, not who said it, so a conversation comes out as one stream of text with timestamps. Speaker labelling needs a separate model this page does not run.
- How accurate is it?
- On clear speech, very: a test recording came back word for word. Background noise, overlapping voices, strong accents and unusual names are where it slips, and the larger models slip less. It can also occasionally repeat a phrase or invent one in a long silence, so treat it as a first draft.