How to Transcribe Audio: 5 Ways (Free and Paid)
You've got a 40-minute recording on your phone: an interview, a lecture, a voice memo you rambled into on a walk. Now you need the words on the page, ideally without paying anyone or losing an afternoon to it.
The quickest free route is to upload the file to an online transcription service. Laxis, for example, takes a recording you upload in its web app and returns a transcript with speakers and timestamps, and a free Laxis account covers 300 minutes a month. Uploading isn't the only way, though, and it isn't always the right one. Below are five honest ways to transcribe audio, what each costs, and when each one wins.
Key points
- A file you already have: upload it to an online transcription service, wait a few minutes, download the text.
- Microsoft 365 or iPhone 12 or later: Word's Transcribe and Voice Memos do it at no extra cost.
- ChatGPT: audio uploads work on paid plans only.
- Meetings and interviews every week: a dedicated app adds speaker labels, summaries and search.
- A short clip or a quote you'll publish: type it yourself, or check every line of an AI draft against the audio.
The 5 ways to transcribe audio at a glance
| Method | Cost | Best for | Limits to know |
|---|---|---|---|
| Online transcription service | Free tier, then a subscription | A recording you already have | Monthly minutes, per-file length and file size |
| Built-in tools (Word, Voice Memos, Pixel Recorder) | Included with Microsoft 365 or your phone | Files on a device you already use | Account, device and language requirements |
| ChatGPT | Paid ChatGPT plan | Short recordings you also want summarized | No audio uploads on Free; size and length limits |
| Dedicated transcription app | Free tier, then roughly $8 to $30 a month | Regular meetings, interviews and lectures | Speaker labels and history on the free plan |
| Manual transcription | Your time | Short clips, poor audio, exact quotes | Slow |
1. Upload the file to an online transcription service
For a recording you already have, this is the path of least effort. You upload the file, the service runs speech recognition, and you download text a few minutes later. Many services work this way; here's how it goes with Laxis:
- Sign up free at laxis.com, or log in.
- Open the Laxis web app, click the arrow next to New Recording and choose Upload Audio File.
- Choose your recording (a meeting, an interview, a lecture or a voice memo) and the language spoken in it, then click Upload.
- Once it's transcribed, the recording appears in your meeting list. Open it and read the transcript, labeled by speaker and timestamped, fixing names and figures as you go.
- Download it as text, Word, PDF or subtitles (.srt), with the summary and timestamps if you want them.
Uploads to the Laxis web app can be m4a, mp3, wav or flac, at most 200 MB each, and Basic, which is free, includes 300 minutes every month. The upload takes audio files only, so pull the soundtrack out of a video first; the conversion tips further down show how.
Other free tiers worth knowing, as of November 16, 2026:
- TurboScribe: three files a day, up to 30 minutes each, one at a time.
- Otter: 300 minutes a month on the free plan, but 30 minutes per conversation and only three file imports in total.
- Notta: 120 minutes a month on the free plan, with each conversation capped at 3 minutes.
The number that trips people up is the per-file limit, which is separate from the monthly total. A tool can offer hundreds of free minutes and still refuse a single 90-minute file, so check that limit before you upload a long recording, not after it fails.
2. Use the tools you already have
Before you sign up for anything, check what's already on your devices.
Microsoft Word (Transcribe). If you pay for Microsoft 365, Word can transcribe an uploaded file: go to Home, open the Dictate drop-down, choose Transcribe, then Upload audio. It accepts .wav, .mp4, .m4a and .mp3, separates speakers as Speaker 1, Speaker 2 and so on, timestamps each section, and works in more than 80 locales. A Microsoft 365 subscription covers up to 300 uploaded minutes a month, and Word keeps the audio in OneDrive, in a folder named Transcribed Files. Availability varies by app and account type, so check that Transcribe appears under Dictate. Word's Dictate button, by contrast, types your live speech; see how dictation works in Microsoft Word for that.
Voice Memos on iPhone. On iPhone 12 or later, Voice Memos transcribes the recordings you make in it, in ten languages. Tap a memo, tap the More button, and choose View Transcript; Copy Transcript takes the whole text. Our guide to transcribing voice memos has the steps and the fixes when no transcript appears.
Recorder on Pixel. On Pixel phones, the Recorder app transcribes as you record.
Google Docs voice typing often comes up in this context, but it types live speech from your microphone; it can't open an audio file, and playing a recording at it through a speaker gives poor results. See our guide to Google Docs voice typing for what it does well.
3. Ask ChatGPT
ChatGPT can transcribe some audio, with limits. According to OpenAI's help center, audio file uploads are available on paid ChatGPT plans and not on the Free plan, and file size and length limits apply. Where ChatGPT shines is the next step: tidying a transcript you paste in, or pulling out decisions and action items.
Two cautions. A model asked to clean up text can drop or reword things, so check quotes, names and numbers against the audio. And look at your data settings (Settings > Data controls) before you upload anything confidential.
We go through plans, file limits and prompts in Can ChatGPT transcribe audio?
4. Use a dedicated transcription app
If you transcribe regularly (weekly meetings, research interviews, lectures), a dedicated app saves you the upload-and-clean-up loop. What you're paying for is everything around the words: speaker labels, summaries, search across recordings, and recording directly in the app.
- Laxis records meetings and in-person conversations in its desktop and phone apps, takes uploads, and returns a speaker-labeled transcript plus a summary and action items; it handles 60+ languages. More on it below.
- Otter and Notta offer similar meeting-and-upload workflows. Their free plans are covered above; Pro plans start at $8.33 and $8.17 a month billed annually.
- Rev sells transcription by the minute: AI transcripts at $0.25 a minute, or human transcription at $1.99 a minute with a 99% accuracy guarantee and delivery in 12 hours or less.
- Whisper, OpenAI's open-source speech recognition model, runs on your own computer under the MIT license. It costs nothing, but setting it up takes some technical comfort.
For interviews in particular, our guide to transcribing an interview covers formatting and quotes.
Why transcripts come out messy, and the fixes that matter
Here's the part that separates a usable transcript from a mess, and it has surprisingly little to do with which tool you picked. Modern engines handle clear audio well. What varies is the audio you give them.
- Background noise. Air conditioning, traffic, a busy coffee shop or wind on a phone microphone. When the noise gets loud relative to the voice, the engine starts guessing.
- Distance from the microphone. A microphone near the speaker captures the full voice; a laptop across the room picks up a thin, echoey version.
- People talking at once. Overlapping speech is hard for any engine to follow, and it's exactly where speaker labels slip.
- Names and jargon. Proper nouns, product names and technical terms are where good tools still stumble. Setting the right language helps, and a custom vocabulary helps more.
Three of those four are fixed before you press record: move the microphone closer, close the window, switch off the fan. Ten seconds of setup saves a long correction pass.
The other common frustration is the shape of the result: one unbroken block of text with no names, no paragraphs and no summary. Speaker labels (diarization) are often the first thing a free tier leaves out, so check for them before you build a habit around a tool.
If you'd rather get the speakers, the timestamps and a summary in one pass, try Laxis free. Even the free plan labels speakers, and new accounts get a 14-day Premium trial with no credit card.
5. Transcribe by hand
Manual transcription still has its place: a two-minute clip, a quote you're going to publish, or audio so poor that machines give up. A few habits make it bearable.
- Slow the playback. Most players can drop to 0.75x or lower, so you pause less often.
- Use keyboard shortcuts to pause and jump back a few seconds without leaving your document. If you do this often, a foot pedal frees your hands entirely.
- Type first, format later. Get the words down, then add speakers, paragraphs and timestamps in a second pass.
- Start from a draft. For anything longer than a few minutes, run an AI transcript first and correct it against the audio. Checking is much faster than typing.
When the file won't upload: convert it first
Often the obstacle isn't transcription at all; it's the file. Any audio converter handles the common cases, and so does ffmpeg if you're comfortable with the command line.
- M4A from iPhone. Voice Memos saves .m4a files. If a tool or a form only accepts MP3, convert the file to an MP3 first.
- Huge WAV files. WAV is uncompressed, at about 10 MB per minute of CD-quality stereo, so a 20-minute recording already reaches the 200 MB Laxis upload limit. Converted to an MP3 at 192 kbps, the same minute takes about 1.4 MB.
- Video files. Laxis uploads take audio only, so pull the soundtrack out first: export the audio from your editing app, or use a converter or ffmpeg. Our guide on how to transcribe a video to text has a one-line ffmpeg command and everything else about video.
For meetings, skip the upload step
If the recordings you transcribe are your own meetings, sales calls or interviews, uploading them one at a time gets old. An AI note taker records the conversation and transcribes it as it happens. Laxis records your Teams, Zoom or Google Meet call from your computer's own audio, so by default nothing joins the meeting; if you prefer, you can turn on a visible assistant bot that does. The phone app handles calls and face-to-face conversations. Every recording returns as a speaker-labeled transcript, with a summary and the action items on top.
What to know before relying on it:
- Processing happens in the cloud, which means you need to be online.
- Basic is free: 300 minutes a month, counted together across meetings and Voice Keyboard dictation. A single recording ends at the 45-minute mark and is saved automatically, and the history view covers 30 days; older items move to an archive instead of being deleted.
- Premium costs $9.99 a month on a yearly plan ($15.99 month to month) and raises the limits to 2,000 minutes a month and recordings of up to 2 hours. CRM sync (HubSpot, Salesforce) comes with the Business plan.
- Laxis isn't HIPAA compliant and signs no BAA, so keep patient information out of it.
Recording other people comes with a duty to tell them, bot or no bot, and Laxis reminds you to let participants know. Get consent where the law requires it. The rules differ from place to place, and nothing here is legal advice; is it illegal to record a conversation? sets them out.
The bottom line
Transcription stopped being the hard part. The engines are good enough that the bottleneck has moved to the few seconds before you record and the choice of tool afterward. For a one-off file, upload it to a service with a free tier. For something already in Word or on your iPhone, use what's built in. For conversations you have every week, record them somewhere the transcript, the speakers and the summary arrive together.
Download Laxis: 14-day Premium trial, no credit card.
Frequently asked questions
How do I transcribe audio to text for free?
The easiest free way to transcribe audio is to upload the file to an online transcription service with a free tier. Laxis transcribes uploads in its web app, with 300 minutes a month on a free account, and TurboScribe allows three 30-minute files a day. If your phone supports it, Voice Memos on iPhone 12 or later transcribes its own recordings at no cost.
Can Microsoft Word transcribe audio?
Yes, Microsoft Word can transcribe audio files if you have a Microsoft 365 subscription. Go to Home, open the Dictate drop-down, choose Transcribe and select Upload audio. Word accepts .wav, .mp4, .m4a and .mp3 files, separates speakers and adds timestamps, and a subscription covers up to 300 uploaded minutes each month.
Can ChatGPT transcribe audio?
ChatGPT can transcribe audio files on paid plans but not on the Free plan, according to OpenAI's help center, and file size and length limits apply. It is most useful for cleaning up or summarizing a transcript you already have. Check names, quotes and numbers against the recording afterward, because a cleanup pass can change the wording.
How do I transcribe a recording with several speakers?
To transcribe a recording with several speakers, use a tool with speaker labels, also called diarization, which tags each line with the person who spoke. Many free tiers leave it out, so check before you upload. Expect labels to slip when people talk over each other, and spot-check the transcript wherever the speaker changes.
What audio formats can I transcribe?
Most transcription tools accept MP3, M4A and WAV, and many also take FLAC. Laxis takes audio uploads only: m4a, mp3, wav or flac files of up to 200 MB. If a file is too large or in the wrong format, convert it with any audio converter or ffmpeg. An MP3 copy of a big WAV is far smaller, and an MP3 works where a form or tool rejects M4A.
How long does it take to transcribe audio?
AI transcription of a typical recording usually takes a few minutes, and rarely longer than the recording itself. Human services take longer: Rev delivers human transcripts in 12 hours or less. Typing a transcript yourself is by far the slowest option, so for anything longer than a short clip, start from an AI draft and correct it.
Related Product
Laxis Meeting Assistant
Automate your notes today.
- Bot-free notes for Zoom, Google Meet, Teams and in-person meetings
- AI summaries, action items and follow-ups in 60+ languages
- Mac, Windows, iPhone and Android, with a free plan
Start for free, no credit card required.
Learn More ▷