A meeting, an interview, a lecture or a voice memo is far easier to search, quote and share as text. This page lets you transcribe audio into text you can copy, save as a plain text file, or save as subtitles with the time of each line.
The audio transcription runs on your own computer, inside this browser tab. Your recording is never uploaded and neither is the text, so it suits private calls and notes too. The first time, your browser downloads the files it needs from our site, about 95 MB, and keeps them for next time.
How to transcribe audio to text
- Press Select a recording and choose a sound file, or drop it anywhere on the page. MP3, WAV, M4A, AAC, OGG, Opus and FLAC all work, as does the sound of an MP4 video.
- Under Language, leave Detect the language, or pick the language spoken. Picking it is a little faster and avoids a wrong guess.
- Press Transcribe. The page reads the recording, gets ready the first time (a one-off download), then works through the recording half a minute at a time.
- Read the text in the box. Press Copy the text, Save as TXT or Save as SRT subtitles.
While it works, a bar shows how far it has got and roughly how long is left. You can use other tabs meanwhile, but keep this one open: closing it stops the work. Press Stop at any time to start again.
If the language was detected wrongly, choose the right one under Wrong language? and press Transcribe again. The recording is already read, so the second run goes straight to the speech.
How long it takes, and the limits
Turning audio to text is hard work for a computer, and this page does all of it on yours. How long it takes depends on your computer far more than on your connection. We timed it on a recent laptop with clear English speech.
| Recording | Time to transcribe | Notes |
|---|---|---|
| 1 minute | About 40 seconds | Measured in Chrome; Firefox was a little quicker |
| 5 minutes | About 3 to 4 minutes | Measured in Chrome and in Firefox |
| 30 minutes | Roughly 20 minutes | Worked out from the 5-minute time; the longest recording the page takes |
- An older or busy computer can take as long as the recording itself, or longer.
- The first time also includes the download of about 95 MB. After that it comes from your browser in a moment.
- Recordings up to 30 minutes long are taken. For anything longer, cut it into parts first and transcribe each one.
- It works best on a computer, in an up-to-date Chrome, Edge or Firefox. On a phone it is slow, and long recordings may not finish.
Getting an accurate audio transcription
The text is usually good for clear speech, but it is never perfect. Names, numbers and specialist words are the most common mistakes, so read it through before you rely on it. These things help most.
- Clear recordings. One person close to the microphone, little background noise and no music under the voice give the best text. A quick microphone test before you record is worth it.
- The right language. English and other widely spoken languages come out best. Pick the language yourself if you know it.
- One language at a time. A recording that switches language half way is written in the language picked or detected.
- Speech, not singing. Song lyrics come out patchy, especially with loud music behind them.
The page does not tell speakers apart, so the text does not say who said what. Each line is one stretch of speech, in the order it was spoken.
When to transcribe audio yourself
Use this page when you want a quick, private draft: notes from a meeting, a rough transcript of an interview to quote from, the words of a voice memo or a lecture to search later. It saves hours of typing, and you only need to correct it.
For a word-perfect transcript, such as for a court, a medical record or a published interview, treat the text as a first draft and check every line against the recording. Nothing written by a machine should be used as is where a mistake matters.
Plain text or SRT subtitles
| File | What is in it | Good for |
|---|---|---|
| TXT | The words, one stretch of speech a line | Notes, minutes, quotes, search, pasting into a document |
| SRT | Numbered subtitles with start and end times, two lines of at most 42 characters each | Captions on a video, or finding a moment in the recording |
Each file takes the recording's name, so team-call.m4a gives team-call.txt and team-call.srt. To burn the subtitles into a video, open it in add subtitles to a video and choose the SRT. If your speech is in a video, the page reads its sound directly, or you can save the sound first with extract audio.
Want MP3 to text? An MP3 is read as it is, and so is a big WAV, so there is no need to convert first. Use WAV to MP3 only if you also want a smaller copy to share.
Private by design: nothing is uploaded
Most transcription services send your recording to their servers. This page does not. The recording is opened, read and turned into text inside this browser tab, and neither the sound nor the text is sent anywhere, to us or to anyone else.
The only thing that crosses the internet is a one-time download of about 95 MB from our own site. It is the same for everyone and holds nothing about you. Your browser keeps it so the next transcript starts quickly, and logging out of Whatafile clears it. Our privacy page has the details.
Turn your recording into text
Choose a recording above, pick its language and press Transcribe. Everything happens on your computer, and you can save the text or subtitles when it is done. More ways to work with sound are in the audio tools.