Transcribe Audio turns recordings into text you can search, quote and edit, without paying per minute and without uploading private conversations. Choose the file and the language, and the transcript appears as it is recognised.
What people transcribe
- Interviews for articles, research and podcasts, to find quotes quickly.
- Meetings and calls recorded on a phone or a computer, for notes and minutes.
- Lectures and classes, to study from text instead of replaying the recording.
- Voice memos with ideas, to-do lists or drafts.
- Voice messages from WhatsApp or Telegram, to read instead of listen.
How to transcribe a recording
- Choose the recording. Phone voice memos (M4A), MP3, WAV and video files all work.
- Choose the language. For English, English (fast) is quick; switch to English (more accurate) if the speakers have strong accents or the recording is noisy.
- Click Transcribe and keep the tab open. You can read the text while it is still working.
- Correct the text in the box, then download it as a text file or click Copy text.
Getting an accurate transcript
| Problem | What helps |
|---|---|
| Fan, traffic or room noise | Clean it with Noise Reduction first, then transcribe the result |
| Accents, fast speech | English (more accurate), or the specific language instead of Detect |
| Wrong language detected | Choose the language yourself |
| Names and technical terms | Correct them in the box; they are the most common errors |
| Long silences | Nothing: silence is skipped automatically |
What to expect
The transcript is a strong first draft. Clear speech close to the microphone comes out with few errors; overlapping voices, music in the background and people far from the microphone cause more. The text is split into lines at pauses in the speech, which makes it easy to read and check. Speakers are not labelled, so add names while you read it through.
Very long recordings are no problem in principle, but take longer and need more memory. On a phone, keep a single recording under an hour; on a computer, several hours work. The speech recognition model is downloaded once and then kept by your browser, so the next transcription starts straight away.
Need subtitles instead? Switch Download as to SRT or see the Subtitle Generator.