Try it without an account

Transcribe an MP3 file

The format everything records in, turned into text with timestamps and speakers. Drop the file and read it in minutes.

Drag your file here

MP3, WAV, M4A, MP4, MOV… up to 10 GB

  • Any bitrate, mono or stereo, any length
  • Language detected on its own, 90+ of them
  • Text, subtitles and summary from one pass
  • MP3

What this tool does

MP3 is what a voice recorder, a phone app or a podcast host gives you, and it's a perfectly good format to transcribe from. Compression throws away what the ear can't hear, and speech recognition doesn't need what the ear can't hear.

Bitrate barely matters either. From 64 kbps upwards the accuracy is essentially the same; what does move it is how close the microphone was and whether two people talk at once. A 32 kbps recording from across a room is the one case worth re-recording.

What we handle that others don't

  • Any bitrate from 32 kbps up, mono or stereo, constant or variable.
  • Long files handled whole: a three-hour podcast doesn't need splitting.
  • The ID3 tags don't matter, and neither does the file name.
  • Speakers separated even when the MP3 is a mono mixdown of several mics.

How it works

  1. 1

    Drop the MP3

    Up to 10 GB — hours of audio in a single file.

  2. 2

    It transcribes itself

    About three minutes per hour of recording, with progress on screen.

  3. 3

    Download what you need

    DOCX, PDF, SRT, CSV or JSON, with your corrections in.

What people use it for

Voice recorders

Almost every handheld recorder writes MP3 by default. Drop it in as it comes.

Podcast episodes

The published MP3 is enough: no need to go back for the session files.

Voice notes

A phone memo of an idea at two in the morning is still a transcribable file.

In detail

Where almost every MP3 here comes from

A voice recorder. And voice recorders record at low bitrates — 32, 48 or 64 kbps — because they're built to fit hours into little memory. That isn't a problem for transcription: at those settings the voice is entirely intact, and what's been dropped are frequencies nobody uses to speak.

The other common origin is a video someone already stripped the audio from, or a downloaded podcast. Both arrive at 128 kbps or better, which is plenty. Compressing further before uploading isn't necessary and saves no quota: minutes are counted by duration, not by size.

Long files and multi-hour recordings

A three-hour MP3 takes little more than 150 MB, so size is rarely the limit. What's worth checking is something else: if the recording starts with half an hour of an empty room, trim it first and save those minutes of quota.

And if there are several separate sessions inside one file, splitting it is better. Not for the system, which will swallow it whole, but because searching inside a three-hour text afterwards is awkward and the summary has to choose between three unrelated subjects.

Questions

And if what you have is the video?

Upload it and you'll get the transcript with timestamps, the speakers separated by name and a summary with the key points. You can start without an account.