How to transcribe an audio recording to text

Quick comparison

1. Microsoft Word's Transcribe feature

If you already pay for Microsoft 365, Word can turn an uploaded recording into a transcript with speaker labels. It works in Word for the web (in Edge or Chrome) and in Word for Microsoft 365 on Windows.

1. Open a new document and sign in with your Microsoft 365 account.
2. Go to Home > Dictate > Transcribe.
3. In the Transcribe pane, choose Upload audio and pick a .wav, .mp4, .m4a or .mp3 file.
4. Keep the pane open while it works. It can take about as long as the recording.
5. Rename "Speaker 1" and "Speaker 2" to real names, then add the transcript to your document.

The catch is the allowance: as of October 2026, Microsoft lists 300 minutes of uploaded audio per month for Microsoft 365 subscribers (more with a Copilot license). Your audio is processed in Microsoft's cloud.

2. The Google Docs voice typing workaround

Google Docs has no "upload audio" button. Its voice typing (Tools > Voice typing, in Chrome) only listens to a live microphone. People get around this by playing the recording out loud next to the mic.

1. Open a Google Doc in Chrome and choose Tools > Voice typing.
2. Click the microphone, then play your recording through speakers near the mic.
3. Stay on the tab until it finishes; voice typing stops if you click away.

It is free, but it runs in real time, misses words when the audio is quiet or muffled, adds no punctuation unless you say it, and cannot tell speakers apart. Use it for a two-minute clip, not an hour-long meeting.

3. Your phone's recorder app

If the recording was made on your phone, the easiest transcript may already be there.

• iPhone Voice Memos: on iPhone 12 or later with iOS 18 or newer, open a recording, tap the More button and choose View Transcript or Copy Transcript. It supports English and several other languages, but it doesn't label speakers. Our guide to transcribing voice memos covers it step by step.
• Google Pixel Recorder: transcribes as you record, and can label speakers on recent Pixels.
• Samsung Voice Recorder: recent Galaxy phones with Galaxy AI can turn a recording into a transcript with speakers separated.

Phone transcripts are great for quick notes. They are harder to work with for long recordings, editing and exporting, and they only help with audio that was recorded or saved on that phone.

4. Online transcription services

Web services like Otter, Rev, Descript and many others let you upload a file and get back a transcript with speaker labels and timestamps. Their strengths are real: many support lots of languages, some offer human transcription for legal or broadcast-grade accuracy, and teams can share and comment on transcripts.

The trade-offs are cost once you pass the free tier, monthly minute caps, and privacy: your recording is uploaded to and stored by another company. That's fine for a podcast episode. Think twice for a confidential interview, a doctor's appointment or an internal strategy meeting.

5. Local apps that transcribe on your computer

Speech recognition models are now good enough, and computers fast enough, that you can transcribe without uploading anything. Local apps download a speech model once and then work offline, with no per-minute fees. Read more about the approach in our guide to local AI transcription.

Transcribing a file with Hark

Hark is a free desktop app for Windows, Mac and Linux. It was built to record meetings, but it transcribes any recording you already have just as well:

1. Download Hark and sign in with a free account (an emailed code, no card).
2. Drag your file onto the window, or choose File > Import. It accepts .m4a, .mp3, .wav, .mp4, .mov, .ogg, .opus, .flac, .webm and more.
3. The first time, Hark downloads its speech models (about 730 MB). After that it works offline.
4. Hark transcribes the file on your computer and works out how many people spoke and who said what.
5. Listen to a short clip of each voice and give it a name. Hark remembers that voice for future recordings.
6. Read through, fix any line by editing it, and export the transcript.

Things to know: Hark transcribes English only, and the first model download takes a while on a slow connection. There are no limits on how much you transcribe, because it all runs on your machine.

6. Typing it yourself

For a very short clip, or audio so noisy that software gives up, typing by hand is still an option. Slow the playback down, use a media player with a keyboard shortcut for pause and rewind, and type in short bursts. Expect it to take several times the length of the recording. Most people now let software do a first draft and spend their time correcting it instead.

How to get a more accurate transcript

Start with better audio

No tool fixes audio it can't hear. If you are recording something you'll transcribe later, put the microphone close to the speakers, turn off fans and notifications, and avoid rooms that echo. For calls, recording each side as its own track (your microphone and your computer's audio) beats holding a phone up to a speaker. Our guide to recording computer audio shows how.

Keep speakers separate

Speaker labels go wrong most often when people talk over each other, sound alike, or say only a word or two at a time. Asking people to take turns helps the transcript as much as it helps the conversation. When labels are wrong, fix them early; see how speaker labels work and how to correct them.

Proofread the parts that matter

Automatic transcripts are drafts. Names, numbers, product names and technical terms are where they slip most. Listen again to anything you'll quote, publish or act on. A tool that plays from the word you click makes this much faster: in Hark, the word being said is highlighted as it plays, and clicking a word jumps there.

Teach the tool your vocabulary

Many tools let you add custom words. With a Hark subscription, Claude proofreads the transcript using your "Names & terms" list to fix misheard names and jargon. That step sends the transcript text (not the audio) to Claude, and every change can be undone.

Privacy: where does your recording go?

Before you pick a tool, ask what is in the recording. A lecture or a podcast is low risk. A job interview, a therapy session, a legal consultation or a board meeting is not. With cloud tools, the audio is uploaded and kept under that company's policies. With local tools, it stays on your computer.

In Hark, recordings, transcripts and voiceprints stay on your computer. Free use sends nothing anywhere. If you turn on the optional AI notes, the transcript text is sent to Claude to write them; Hark doesn't train AI on your recordings.

If the recording includes other people, remember that recording laws vary by place, and some require everyone's consent. Get permission before you record.

Which method should you use?

• A quick memo you recorded on your phone: use the phone's own transcript.
• An occasional file and you already pay for Microsoft 365: Word's Transcribe.
• Non-English audio, or you need human-verified accuracy: an online service.
• Private, long or frequent recordings in English: a local app like Hark.

Working with a specific format? See MP3 to text and MP4 to text.