How to Transcribe an Audio File to Text in 2026: MP3, M4A, WAV

To turn an audio file into a transcript, upload it to a transcription tool and export the text. With Microsoft 365, use Transcribe in Word for 300 free minutes a month. Otter's free plan takes three imported files. On a Mac, MacWhisper transcribes MP3, M4A and WAV files offline. For a perfect record, pay for a human transcript.

The file is usually sitting there already: a customer interview saved as an M4A, a recorded lecture, a voice memo from the car, a podcast episode someone sent over. Listening back to find one sentence takes forever. This guide walks through four ways to get the words out, step by step, what each one costs and where your audio goes, plus how to fix a file that won't upload or comes back full of mistakes. If you're still choosing software, the free transcription software roundup compares eight tools side by side.

Which way to transcribe your audio file

The right method depends on three things: how long the file is, whether it can leave your computer, and how exact the words need to be. The limits in this table are taken from each maker's pricing or help pages, which are linked in the sections that follow.

One thing to know about Apple's built-in apps: Notes on a Mac transcribes audio, but Apple's Notes guide only covers recordings made inside Notes, not files you already have. For an existing MP3, use one of the methods below.

How to transcribe an audio file in Microsoft Word

Microsoft Word's Transcribe tool turns an uploaded MP3, M4A, WAV or MP4 file into a transcript split into sections by speaker. Microsoft's Transcribe guide gives Microsoft 365 subscribers 300 minutes of uploaded audio a month, and says Word for the web needs the new Microsoft Edge or Chrome and an internet connection.

1. Sign in to Microsoft 365 and open a new document in Word for the web.
2. Go to Home, open the Dictate dropdown and choose Transcribe.
3. In the Transcribe pane, select Upload audio and pick your file.
4. Wait for the transcript to appear in the pane. Longer files take longer.
5. Rename speakers: hover over a section, select Edit transcript section, change the label, tick Change all Speaker 1 (or whichever number) and select Confirm.
6. Select Add all to document to drop the whole transcript into the page, or Add section to document for one part.

Microsoft stores the audio in a Transcribed Files folder in OneDrive, where you can rename, share or delete it. If the Transcribe button is missing, Microsoft's page says to check that you're signed in with an active Microsoft 365 subscription. The same guide says Transcribe in Word on Windows is limited to business (commercial) accounts, so on a personal plan, follow the Word for the web steps above.

Word is where I start when the transcript is going into a document anyway. When I had a recorded customer interview to turn into a write-up for my team, uploading it straight into the Word file meant the quotes landed next to my notes, and fixing "Speaker 2" to the customer's role took one edit for the whole transcript.

How to transcribe an MP3 with Otter's free plan

Otter transcribes imported audio files with speaker labels, and the free Basic plan is enough to try it on a couple of files. Otter's pricing page caps Basic at 3 audio or video file imports per account, ever, along with 300 minutes a month and 30 minutes per conversation. Pro raises that to 10 imports a month and up to 90 minutes per conversation, at $16.99 monthly, or $8.33 per month with annual billing.

1. Sign in at otter.ai.
2. Choose Import, or drag the file onto your Conversations list, as Otter's MP3 page describes.
3. Wait for the transcript, then click any word to jump to that point in the audio.
4. Tag speakers and correct names in the transcript.
5. Export the text. Basic exports as TXT; Pro adds PDF, DOCX and SRT, per the pricing page.

How to transcribe audio on a Mac without uploading it

MacWhisper transcribes audio files on your Mac using local models, so the recording never goes to a server. The MacWhisper site lists support for MP3, WAV, M4A, OGG, Opus, MOV and MP4 files, exports to TXT, DOCX, PDF, Markdown, SRT and VTT, and a free version, with Pro as a one-time €64 license.

1. Download MacWhisper from its site, or get it as Whisper Transcription on the Mac App Store.
2. Open the app and drag your audio file onto the window.
3. Pick a model. Smaller models are faster; larger ones are more accurate and slower.
4. Start the transcription and leave it running. Speed depends on your Mac and the file length.
5. Read through, fix names, then export in the format you need.

This is the route for anything confidential: an HR conversation, a board discussion, a medical appointment you recorded with permission. When the file holds something I wouldn't paste into a web form, I'd rather wait a few extra minutes for my own Mac to do the work.

The free command-line route: Whisper

If you're comfortable in Terminal, OpenAI's Whisper repository holds the open-source speech model MacWhisper is named after, free and released under the MIT license. It needs Python and the ffmpeg tool. On a Mac with Homebrew, setup and a first transcript look like this:

brew install ffmpeg
pip install -U openai-whisper
whisper interview.mp3 --model turbo --language English

By default Whisper saves the transcript as plain text plus SRT and VTT caption files with timestamps, according to its command-line options. The repository doesn't mention speaker labels, so for a two-person interview you'll mark who's talking yourself.

How to get a transcript from Gemini

The Gemini app accepts audio uploads, which makes it a quick option for a short clip when you also want a summary. Google's file upload guide caps free users at 10 minutes of total audio; Google AI Pro and Google AI Ultra extend that to 3 hours, with up to 10 files per prompt.

1. Open the Gemini app or gemini.google.com.
2. Use the upload option in the prompt box to add your audio file.
3. Paste a prompt that asks for the exact words, not a tidy rewrite (a template is below).
4. Copy the transcript into a document and check it against the audio.

Transcribe this audio word for word. Don't summarize or tidy the wording.
Label each speaker (Speaker 1, Speaker 2) and start a new line when the speaker changes.
Add a timestamp every few minutes.
These names and terms come up: [list names, company names, product terms].
Mark anything you can't hear clearly as [inaudible].

A chat assistant is built to be helpful, which sometimes means smoothing over a stumble or a half-finished sentence. Asking for word-for-word output helps, but I still spot-check any line I plan to quote. Google's Gemini privacy page also says a subset of chats is read by human reviewers and asks you not to enter confidential information you wouldn't want a reviewer to see; turning off Keep Activity stops future chats being reviewed.

When to pay for a human transcript

Pay for a human transcript when every word has to be right: a legal matter, academic research, a published interview or anything someone might dispute. Rev's pricing page lists human transcription at $1.99 a minute, with a 99%+ accuracy guarantee and turnaround within 12 hours. Its free plan includes 45 minutes of AI transcription a month if you want to try the automatic version first.

A 30-minute recording at that rate comes to about $60, so for most work meetings an automatic transcript and ten minutes of cleanup is the better trade. For a quote going into a press release or a contract dispute, the $60 is cheap.

How to get a cleaner transcript from any tool

Most transcript errors come from the audio and the names, not the tool. A few minutes of preparation saves a lot of correcting afterward.

• Start with the best copy of the file. A recording passed along through a messaging app can arrive compressed; ask for the original if you can.
• Cut the dead air. Long silences at the start and end waste minutes on capped plans like Otter's and Gemini's free tiers.
• Write down the names first. Before uploading, I list everyone who speaks and the product or company names that will come up, then use that list to fix the transcript with find and replace.
• Check the numbers by ear. Prices, dates and percentages are where a transcript does the most damage when it's wrong, so I listen back to each one.
• Keep the timestamps. They're how you find the exact moment again when someone asks "did they really say that?"

The last time a client asked me for an exact line from a recorded call, the transcript got me to the right minute, and the audio settled the wording. That's the right split of jobs: the transcript for finding, the recording for proof.

Why an audio file won't transcribe

When an upload fails or the transcript comes back short, the cause is usually the format, the length or the account.

The file format isn't accepted

Word takes .wav, .mp4, .m4a and .mp3, per Microsoft's guide. If you have something else (an .opus voice note, an .amr phone recording, a .webm export), convert it to MP3 with the free ffmpeg tool, or open it in MacWhisper, which reads OGG and Opus directly.

ffmpeg -i voice-note.opus voice-note.mp3

The file is too long

Free plans cut off long recordings: 30 minutes per conversation on Otter Basic, 10 minutes of total audio in Gemini's free plan. Split the file into parts with ffmpeg's segment option, transcribe each part, and paste the results together in order. This example cuts a file into 25-minute pieces:

ffmpeg -i lecture.mp3 -f segment -segment_time 1500 -c copy lecture-part%02d.mp3

The Transcribe button is missing in Word

Microsoft says to make sure you're signed in with an active Microsoft 365 subscription and using the new Edge or Chrome. Safari and Firefox aren't on its list for Word for the web.

The transcript is full of mistakes

Heavy background noise, people talking over each other and a phone held far from the speaker all hurt accuracy. Try a larger model in MacWhisper, set the language explicitly (Whisper's --language option does this), and if the recording is truly rough, a human transcriber is the honest answer.

For video files rather than audio, most of the same tools work, since Word and MacWhisper both take MP4. If the video is on YouTube, there's a quicker route in the guide to getting a YouTube transcript.

Next time, record with the transcript in mind

Transcribing a file after the fact is the slow way. For meetings you haven't had yet, a tool that transcribes as the call happens saves the upload step entirely; the live transcription guide covers what the big video call apps have built in, and the Otter alternatives page compares note takers by their limits. Whatever you use, tell people you're recording before you start.