How to transcribe audio to text for free
Typing out an interview, lecture or voice note by hand takes hours. A speech model can produce a first draft in minutes. With this tool the model runs on your own device, so your recording is never uploaded.
Transcribe a recording
- Open Speech to Text.
- Add an MP3, WAV, M4A, OGG or a short video file.
- Press Transcribe. The first time, the speech model downloads (about 77 MB).
- Read through the transcript, fix any mistakes, then copy it or download it as a .txt file.
Recordings of up to one hour are supported. For something longer, cut it into parts with the Audio Trimmer.
Getting a more accurate result
- Record close to the speaker, in a quiet room.
- One person speaking at a time works far better than overlapping voices.
- Names, jargon and strong accents often need a quick manual check.
- The language is detected automatically, and common languages work best.
From a video file
Short videos can be added directly. For a large video, extract the soundtrack first with Video to MP3, which gives a smaller file to work with.
Plan before you start
The first time you use the tool, a speech model of about 77 MB downloads to your browser. On mobile data that is worth knowing, so start on Wi-Fi. After that, the transcription itself runs on your own device and the recording is never uploaded. Keep the page open until it finishes, since closing it stops the work.
A good proofreading routine
- Read the whole transcript once while listening to the recording at normal speed.
- Fix names, places, numbers and technical words first, since these are the most likely to be wrong.
- Add paragraph breaks where the speaker changes topic, so the text is easy to read.
- Add speaker names yourself if there is more than one person, since the transcript is plain text.
- Read it once more without the audio to catch anything that does not make sense.
Long recordings
Recordings of up to one hour are supported. For something longer, cut it into parts with the Audio Trimmer and transcribe each part, then join the text. Smaller parts are also easier to proofread, because you can check each one against the audio while it is fresh.
Recording people
Ask for permission before you record a conversation, and tell people how you will use the recording and the transcript. Rules about recording differ between places, and clear consent avoids problems.
Frequently asked questions
Is the transcript perfect?
No automatic transcript is. Treat it as a strong first draft and proofread important parts.
Is my recording uploaded?
No. The model runs on your device. Only the first-time model download uses the internet.
How long does it take?
It depends on your device and the length of the recording. Longer files take longer, and you can leave the page open while it works.
Does it label who is speaking?
No. The result is plain text, so add speaker names yourself while you proofread.
What can I do if the transcript is full of mistakes?
Check the recording quality. Noise, echo and overlapping voices cause most errors. A cleaner recording gives a far better result.
Can I use it for a video file?
Short video files can be added directly. For a large video, extract the audio with Video to MP3 first.