How can I transcribe audio to text for free?

Yes. There are three free routes: a built-in tool, open-source software on your own computer, and the free allowance of a cloud service. The first two cost nothing per recording but use your computer and your time. Cloud allowances are capped by the hour or the minute, so they suit a few recordings rather than a steady flow. Choose the route by how long the audio is, how private it is, and what your computer can run. The guide to transcribing audio to text covers the whole process, and the transcription overview lists the other topics.

Apple's Voice Memos app is free on a Mac that qualifies

If you have a Mac, start with Voice Memos. The app can recognize and transcribe the speech in audio recordings ¹. Nothing extra needs to be installed, and it costs nothing. The feature is not available in every country or region ², so check that it works where you live. If it is not available to you, use one of the routes below.

To take the text out, select it, Control-click it and choose Copy, then paste it into your document.

Open-source Whisper costs nothing to run, but it needs a capable computer

Whisper's code and model weights are released under the MIT License ³. You can run it on your own computer without paying for the software. The cost is the hardware. The largest model needs about 10 GB of video memory ⁴, so a computer with less memory has to use a smaller model, which runs faster and is less accurate. The guide compares the model sizes in a table.

Running Whisper yourself is the right choice when the recording must stay on your machine. The question on offline transcription explains how to confirm that a tool works without a connection.

Free cloud allowances differ in size and in their terms

Cloud services also offer free allowances. They are trials or monthly quotas, and the limit decides whether a recording fits. Check the limit on the day you start, because these pages change.

A single hour-long recording fits inside the monthly allowance of the services above that offer minutes or hours. A backlog of interviews does not. A cloud service also receives the file, so the privacy question in the guide applies to every cloud route, free or paid.

A free tool still leaves a draft that needs proofreading

Free software gives you a draft, not a finished record. The usual measure of how many words a tool gets wrong is the word error rate, and a low rate still leaves some errors to find. Plan time for a read against the audio, especially for names and numbers. The comparison of human and AI transcription explains when a person is worth paying for.

Pick the free route by length, privacy and hardware

1. Short recording, Mac that qualifies: use Voice Memos.
2. Private recording, capable computer: run Whisper or an app built on it on your own machine.
3. Long recording that is not sensitive: use a cloud allowance, after you read its data terms.
4. No Mac and no capable computer: use a free file transcription page in your browser, such as our audio-to-text page. Yap, which we make, is an on-device dictation app for Mac and Windows. It is built for dictating into the app you are using, not for transcribing a finished file, so use the browser page or the tools above for files.

Before you start, check that your file format is accepted. The question on audio formats lists which ones each route takes. If the recording is quiet or noisy, read the question on low-quality audio first, because a poor recording lowers accuracy whatever the price.