Skip to content
FlexiConvert

One file at a time, up to 100 MB.

Choose the language being spoken, so the words are recognized and spelled correctly.

Your files are deleted automatically after conversion, and no later than one hour after upload. How we handle your files

Advertisement

Turn an MP3 into text without typing it out by hand. Upload an interview, a lecture or a podcast episode, choose the language being spoken and download a .txt with everything that was said, split into paragraphs. It's free, with no sign-up and no watermarks.

How to convert MP3 to text

  1. Select your MP3 file (one at a time, up to 10 minutes long).
  2. Choose the audio language; English is the default.
  3. Press “Convert to text” and wait: it usually takes less than a minute.
  4. Download the .txt and check it before you use it.

Advertisement

What you get

A text file (.txt, UTF-8) with what's said in the MP3, including punctuation and capital letters. A long pause starts a new paragraph, so an interview doesn't come out as one solid block. It opens in Notepad, TextEdit or Word, and you can copy and paste it wherever you need it.

It doesn't tell speakers apart: with two voices, their sentences run on without a name for each one. It has no timestamps either; if you need them for captions, use Video to SRT.

Which recordings work best

Speech recognition does far better on clean sound. In our tests, clear speech in English got about 92% of words right, and clear Spanish more than 97%. Background noise lowers that, and a meeting recorded with one microphone in the room got nearly half of its words wrong.

  • Works well: one person speaking clearly near the microphone, such as dictation, a voice note or a podcast.
  • Hit and miss: interviews in a café, people talking over each other, very strong accents.
  • Works badly: music over the voice, meetings far from the microphone, and songs.

With very poor audio, the AI can even make up a sentence that sounds right but that nobody said. So listen back to the parts that matter, and don't treat the text as a verbatim quote, or use it for minutes or legal documents, without checking it.

Advertisement

Choose the right language

Pick the language being spoken, not the language of this page: words are recognized better and spelled properly. You can choose from 24 languages. If the audio mixes several, choose the one you hear most.

Limits

  • One MP3 at a time, up to 10 minutes and 100 MB. If it's longer, split it and upload the parts one by one.
  • Up to 6 files per hour, plus a daily allowance shared by all users. If it runs out, the tool comes back the next day.
  • The audio goes through Cloudflare for speech recognition, without metadata and without your IP address. It isn't kept or used to train models. There's more in our privacy policy.
  • If your file is in another format, try M4A to text, WAV to text or MP4 to text.

Frequently asked questions

How do I turn an MP3 into a Word document?

Download the .txt here and open it in Word: the text comes in with its paragraphs and you can save it as a document. Word can also transcribe through its Dictate feature, but only with a Microsoft 365 subscription.

How accurate is MP3 transcription?

It depends mainly on the sound. With one clear voice it gets most words right; with noise, several voices or a poor recording it makes quite a few mistakes. Always check the text, especially names and figures.

Can it transcribe an interview with two people?

Yes, but both people’s sentences run on together, with no label for who said what. Long pauses start a new paragraph, which helps you separate them when you check it.

What can I do if my MP3 is longer than 10 minutes?

Cut it into 10-minute parts with an audio editor such as Audacity and upload them one at a time. We set this limit so the service can stay free for everyone.

Why is there a sentence that nobody said?

It’s a known flaw in AI speech recognition: with noise or long silences, it sometimes fills the gap with natural-sounding text. It’s rare with clear speech, but it’s worth checking anything important against the audio.

Advertisement

Advertisement

Advertisement