How to
- Drop an audio or video file.
- Keep automatic language detection and pick a mode: Fast, Balanced or Accurate.
- Transcribe, correct any word in place, then copy the text or download TXT, SRT or VTT.
Drop the audio recording above and you get its text in a few seconds to a few minutes. Speech recognition runs in your browser: the audio is never uploaded, there is no length limit and no account.
What you get
- This is speech recognition, not a format change: it listens to the recording and writes down what is said, with punctuation and capitals.
- The text is split into paragraphs at the pauses. Timestamps are kept on screen, and in the SRT and VTT exports.
- Around a hundred languages, detected automatically. It can also translate the speech into English.
- Clear speech comes out close to perfect. Background noise, people talking over each other and rare names cause mistakes, which you can correct in place before downloading.
- Speakers are not labelled: the text says what was said, not who said it.
What it is for
Use plain text for notes, minutes, quotes from an interview, a voice message you cannot listen to right now, or anything you want to search, copy or translate.
Because nothing is uploaded, it is suitable for confidential recordings: medical, legal, HR interviews, or private voice notes.
Frequently asked questions
Is my audio file uploaded?
No. Speech recognition is downloaded to your browser once and runs there. The recording itself never leaves your device: you can go offline once it has loaded and it still works.
How accurate is it?
On a clear voice, Accurate makes very few mistakes; Fast makes more but runs on any phone, and Balanced sits in between. Every line can be corrected before you download.
How long does it take?
On a recent computer, a minute of speech takes a few seconds with Fast or Accurate, and a little longer with Balanced. Phones are slower. Silences are skipped, so a long recording with pauses goes faster than its length suggests.
Why is there a download the first time?
Recognising speech on your device means downloading the recognition itself: 80 MB for Fast, 245 MB for Balanced, 545 MB for Accurate. It is stored in your browser, so later transcriptions start at once.
Which languages are supported?
About a hundred, including English, French, Spanish, German, Italian, Portuguese, Dutch, Arabic, Chinese, Japanese and Russian. The language is detected automatically; you can also set it yourself.
Is it free? Is there a length limit?
Free, with no account and no limit on minutes: your device does the work, so there is nothing to meter. Very long recordings are limited only by your device's memory.