How do you transcribe an audio file on Windows?

Use Word's Transcribe pane if you have a qualifying Microsoft 365 account and an internet connection. Use Buzz if the recording must stay on your PC or you have no connection. The whisper.cpp command line also works offline, but its example needs a 16-bit WAV file. The route you pick depends on three things: whether you have Microsoft 365 through work, whether you can go online, and whether you are comfortable with the command line.

Word's Transcribe pane is the built-in route, but it needs the internet

Word for Microsoft 365 on Windows has a Transcribe pane that turns speech in an audio file into a text transcript. Microsoft says the feature is currently available only in Word for Microsoft 365 on Windows in Commercial Tenants ¹. Its system requirements list an internet connection ², so it will not run on a plane or in a dead zone.

To transcribe a file that is already on your PC:

1. Open Word and select Home, then the Dictate dropdown, then Transcribe.
2. In the Transcribe pane, select Upload audio, then choose the file from the picker.
3. Wait for the transcript to finish, then review and edit it in the pane.

There are limits. Users with a Microsoft 365 subscription can transcribe a maximum of 300 minutes of uploaded audio per month ³. The audio file, whether recorded or uploaded, is saved to the Transcribed Files folder in OneDrive ⁴, so check your privacy settings before you upload anything sensitive.

Buzz transcribes files on your own PC, with no connection needed

Buzz describes itself as transcribing and translating audio offline on a personal computer ⁵. It can import audio and video files and export the transcript as TXT, SRT or VTT files ⁶. SRT and VTT are subtitle formats, so Buzz is useful if you need captions as well as text.

Buzz is the practical offline choice for a recording you already have. Install it, import the file, and choose the export format. Because the transcription runs on your PC, it does not need a server to do the work ⁵. Read the app's privacy policy and release notes before you trust any claim about network use, and test it with Wi-Fi off.

The whisper.cpp command line needs a 16-bit WAV file

Whisper.cpp, an open implementation of OpenAI's Whisper model, runs from the command line. Its command-line example currently runs only with 16-bit WAV files, so convert your recording first if it is in another format ⁷. A tool such as ffmpeg can do the conversion. This route suits people who are comfortable in a terminal. For the setup steps, see how to install Whisper on Windows.

Cloud services are quick, but they need a connection and a plan

Cloud transcription services run the recognition on their servers, so they need a connection and send your audio to the provider. Amazon Transcribe's free tier covers up to 60 audio minutes a month, but only for the first 12 months after signup ⁸. A free tier is a limit, not a permanent free service, so check the terms before you build a routine on it. For how to think about privacy, see whether voice to text is safe.

Compare the routes before you choose

If you transcribe interviews or lectures for a living, check accuracy before you rely on any route. Our question on how accurate AI transcription is explains how to test it, and the word error rate entry explains the score.

What to do

1. If you have a Microsoft 365 commercial account and a connection, start with Home, then Dictate, then Transcribe in Word.
2. If the recording must stay on your PC, install Buzz and test it with Wi-Fi off on a short file first.
3. If you want to use the whisper.cpp command line, convert the file to 16-bit WAV before you run it.
4. Read the transcript against the recording. Fix names, numbers and any words the app invented, because no route is error-free.

For more on each route, see how to transcribe audio to text, how to transcribe audio offline and whether Word can transcribe audio. The Dictation on Windows overview covers the rest of the Windows options, and our guide to offline speech to text on Windows covers the offline routes in more depth.