Illustration of a sound wave flowing into blank text lines on a laptop screen
Audio to text4 min read

How to transcribe audio to text on your computer

Three ways to turn a recording into text, and the steps to do it on your own computer with DijiFlow Dictate, where the file stays on your drive.

You have a recording, perhaps an interview, a lecture or a meeting, and you need the words as text. This guide shows how to transcribe audio to text on your own computer with DijiFlow Dictate, so the file stays on your drive and there are no minutes to buy.

Three ways to transcribe audio to text

  • Type it yourself. It costs nothing, but listening, pausing and typing takes far longer than the recording itself.
  • Upload it to an online service. Quick to start, but the file leaves your computer, and some services cap long recordings or charge by the minute.
  • Use an app on your computer. The recording is read from your drive and the transcript is written back there. This is how DijiFlow Dictate works, on Mac and on Windows.

How to transcribe an audio file in DijiFlow Dictate

The labels below are the ones the English Mac app shows. If this is your first time with the app, the getting started guide covers the install and the first model download.

  1. Open the Transcribe File tab. Click the DijiFlow Dictate icon in the menu bar, choose Settings, then the Transcribe File tab.
  2. Add the file. Drag the audio file into the window, or click Select a File... in the Select Video or Audio File card. You can pick several files at once; they run one after another.
  3. Set the spoken language. Choose the language of the recording under Spoken language. Naming it gives steadier results than leaving the app to detect it.
  4. Choose Transcript. In the Output format card, choose Transcript. Leave Output language on Same as spoken, unless you want the text in another language (translation is part of Pro and the trial).
  5. Press Transcribe. When the run ends, the text appears in the result box. Use Copy, or Save As... to keep it as a plain text file.

The file transcription page shows the same three-step flow, with the subtitle and translation options next to it.

Long recordings

On Pro and the 30-day trial, a file can be any length. A 90-minute film stays at around 1 GB of memory while it runs, so a feature-length recording works on the machine you already have.

The free version transcribes the first 3 minutes of each file, with a smaller model and on the processor only, so it is slower. There is no limit on how many files you run. The 30-day trial is the full test: it takes files of any length and lets you pick the largest model.

Where the audio goes

The recording is read from your own drive, transcribed on your computer, and the transcript is written back there. Your audio and text are never uploaded. The free version and the trial need no account and no card. The speech model is downloaded the first time you use the app, and larger models when you first pick them.

That matters for recordings you would not hand to a stranger: client calls, family recordings, interviews that are not yet published.

Getting a cleaner transcript

  • Name the language. DijiFlow covers 29 languages in full and recognises 90+ with auto-detect. Outside the 29, languages are not tuned, and the text may come out in English.
  • Use the largest model for hard audio. On Pro and the trial you can pick Large. It takes longer, and it gets more words right in noise and music.
  • Add your names and terms. On Pro and the trial, the Vocabulary tab has Custom Terms: add the names, brands and jargon in your recordings, one per line, and the transcript is corrected towards them.
  • Turn on timestamps or speakers. On Pro and the trial, the Output card has Show Timestamps and Identify Speakers, which labels lines Speaker 1, Speaker 2 and so on.
  • Record cleanly. On clear speech, accuracy is about 98% with the largest model. Distance from the microphone, crosstalk and background noise all lower it.

Whatever the settings, read the transcript before you rely on it, especially names and numbers.

What it does not do

DijiFlow Dictate works on files that are on your computer. It does not take a web link, so a recording on a website has to be saved to your drive first, where you have the right to do so. It is also not a meeting bot: it transcribes the recording after the call, not the call while it happens.

Questions

Which audio formats work?

Most common formats work, including MP3, WAV, M4A, AAC, AIFF and FLAC, and the soundtrack of video files such as MP4 and MOV. If a file is not an audio or video file, the app skips it and says it is not a supported format. Try your own files in the free version first.

Can I transcribe audio to text for free?

Yes, within a limit. The free version is free forever, with no account and no card, and it transcribes the first 3 minutes of each file. For full-length recordings, the 30-day trial unlocks every Pro feature. On a Mac it starts on the app's first screen with Continue with free trial, and needs no card and no email.

Can it turn an MP3 into text?

Yes. Drop the MP3 into the Transcribe File tab and follow the steps above. WAV and M4A files work the same way.

Does it work on Windows?

Yes. DijiFlow Dictate runs on Windows 10 and 11, and on macOS 14 or later with Apple Silicon. The labels in this guide are from the Mac app.

To try it on one of your own recordings, download DijiFlow Dictate. The free version covers the first 3 minutes of each file, and the trial covers whole files.

DijiFlow DictateDijiFlow Dictate

The DijiFlow Dictate Team

Notes on private, on-device dictation and getting more done with your voice.