To transcribe audio to text, upload the file to Supertranscription. A few minutes later you get the text with punctuation, the speakers told apart and timing for each spoken word, ready to copy. Your first file, up to 30 minutes, is free and needs no account. Below: the steps, the formats that work, what really changes the quality, and when this is the wrong tool.

Which files you can upload

The site accepts MP3, M4A, WAV, OGG, FLAC, AAC, WebM, MP4 and MOV. Most recordings are already one of these:

  • the iPhone Voice Memos app saves M4A (the iPhone walkthrough is here);
  • digital voice recorders usually produce MP3 or WAV;
  • Zoom saves a meeting on your computer as an MP4 video and an M4A audio file;
  • a video shot on a phone is MP4 or MOV.

You don't need to pull the sound out of a video: an MP4 or MOV works as it is. A file can be up to 1 GB. A long recording doesn't need cutting either: your browser splits it into parts by itself, and you upload the whole file once. If you only have one format in mind, there are short pages for MP3 to text and M4A to text.

If your file is MKV, AVI, WMA or AMR, convert it to MP3 first, for example with the free VLC player (Media menu, then Convert / Save). A WhatsApp voice message is an .opus file: renaming it to .ogg usually works, and if not, convert it to MP3 too.

The steps

  1. Open supertranscription.com on your computer or phone. The site transcribes English, so there is nothing to choose.
  2. Drag the file into the box, or click it to browse your folders.
  3. Wait until it finishes. Depending on the length, it takes a few minutes. Keep the tab open.
  4. Read the text. It arrives with punctuation, separate turns for each voice (Speaker 1, Speaker 2…) and a time marker for each word.
  5. Take it with you: Copy puts the text on your clipboard with the speaker labels, and Download gives you a text file (.txt) or subtitles (.srt, .vtt).

Without an account, the transcript isn't kept: it exists only in your browser tab. Copy or download it before closing the page. If you are signed in, it is saved in My notes, where you can find it later.

The result is plain text: paste it into a new document to keep working on it.

What it costs

  • Without an account: one file per person, up to 30 minutes of audio. With a longer recording, you get the transcript of the first 30 minutes.
  • With a Google or Apple account: Pro is free until January 2027: 10 hours of transcription a month, no card needed. Nothing is for sale at the moment.

The details are on the pricing page.

What improves the result

The quality of a transcript depends first on the recording itself:

  • Get the microphone close. The phone or recorder should be near the person talking, not at the far end of the table. In a small meeting a phone in the middle can be enough. In a big room it isn't.
  • Look for quiet. Espresso machines, traffic and background music all cause mistakes. For an interview, a busy café is the worst place.
  • Avoid talking over each other. Overlapping voices are hard to separate, and that is where the tool slips most easily, including on who is speaking.
  • Check names and technical terms. Proper names, acronyms and specialist vocabulary are the words most likely to come out wrong. There is no custom dictionary, so fix them in one pass with find and replace in your word processor.

What it's for, and what it doesn't do

A transcript is useful for a journalist's or researcher's interview, a lecture, a meeting or call you want in writing, or subtitles for a video (SRT or VTT). But:

  • No live transcription. The site processes a finished recording. It doesn't write while you speak.
  • No summary on the homepage. Summaries, to-do lists and translation exist only in My notes, through Ask AI (an example for meetings).
  • No translation. The homepage transcribes English only.
  • No speaker names. Voices are numbered. Adding the names is up to you.
  • No legal standing. If every word counts (a hearing, a deposition, an official document), have a person check the transcript word by word, or hire a professional transcriber.

Two free alternatives

  • Voice typing in Google Docs. In the Tools menu, "Voice typing" works in English (Google's help). It listens to the microphone and doesn't read audio files: good for dictating a short text, not for transcribing a recording you already have.
  • Whisper on your own computer. If the recording must not leave your machine (a confidential source, health data), OpenAI's open-source Whisper model runs locally. In return you have to install it (Python, the command line, ffmpeg), and a long recording needs a powerful computer.

Where the recording goes

Speech recognition is done by ElevenLabs (United States), which receives the recording only to process it. We don't keep the audio file uploaded from the homepage. If you record or import audio in My notes, the recording and its transcript stay in your account until you delete them. It is all spelled out in the privacy policy.

Recording another person usually needs their consent, and the rules differ by country and, in the US, by state. Some require only one party to agree, others require everyone on the recording. Tell people before you press the button.

To try it, drop a recording on the homepage. The first one is free.