You can transcribe audio to text for free, with a tool you probably already own. What separates the tools is something else: where the audio goes while it is transcribed.
On a clean recording, automatic tools are increasingly alike. They differ on three points: how much is free, how much proofreading they leave you, and which server listens to your file. This guide covers each method, with steps for the most requested tools.
Which tool to pick in thirty seconds
Answer the questions in order. The first "yes" gives you the tool.
- Is the audio confidential: a client, a patient, an HR interview?yes : Whisper on your computer, or transcribe by hand. Nothing leaves your machine.
- Is it a WhatsApp voice message?yes : WhatsApp’s built-in transcripts, switched on in settings.
- Was it recorded with your phone?yes : Voice Memos on iPhone, or Voice Recorder on Samsung.
- Does your company have Microsoft 365?yes : Word’s Transcribe feature: 300 minutes a month included.
- None of the aboveyes : An online service, with a file that is not sensitive.
The methods, side by side
| Method | Cost | Where the audio goes | Best for |
|---|---|---|---|
| By hand | Free, but 4 to 6 hours of work per hour of audio | Nowhere: it stays with you | Short extracts, exact quotes, highly confidential audio |
| Your phone's app | Included | On the device or with the manufacturer, depending on model and settings | Voice notes, ideas, short interviews |
| Included | On the phone, according to WhatsApp | Voice messages you receive | |
| Word (Transcribe) | Included in Microsoft 365, 300 min of uploaded audio per month | Microsoft servers | Companies already on Microsoft 365 |
| Whisper, locally | Free (open source) | Your computer | Large volumes, sensitive audio |
| Online service | Limited free tier, then a subscription | The service's servers, often outside the EU | One-off transcription of a non-sensitive file |
| AI voice recorder | The device, plus a subscription depending on use | The vendor's servers | Frequent meetings and appointments, with a summary |
Transcribing audio in Word
Word for Microsoft 365 transcribes an audio file and splits the text by speaker, with timestamps. In a company it is often the simplest option: the tool is already paid for, and the audio stays in an environment your IT team knows.
- Open a document in Word for Microsoft 365, on Windows or in the browser.
- On the Home tab, open the Dictate button's menu and choose Transcribe.
- Click Upload audio and select your file (MP3, WAV, M4A or MP4).
- Wait: processing takes a fraction of the file's length.
- Rename the speakers ("Speaker 1" becomes "Claire"), then fix doubtful passages by replaying them straight from the pane.
- Click Add to document, with or without names and timestamps.
The exact position of the buttons varies with the Word version. A Microsoft 365 subscription covers 300 minutes of uploaded audio per month; the feature is for subscribers only.
Good to know. Word accepts the M4A files an iPhone produces. For a format it refuses, the free VLC player converts to MP3 through its Convert menu.
Transcribing a WhatsApp voice message
WhatsApp transcribes the voice messages you receive itself. The feature is off by default.
- Open WhatsApp settings, then Chats.
- Turn on voice message transcripts and choose the language.
- In a chat, long-press a voice message, then tap Transcribe.
According to WhatsApp, the transcript is computed on your phone and only you see it: the sender does not know you read their message instead of listening to it. Language support is wider on iPhone than on Android; check your settings.
If your language is not available, the voice message can be shared as an audio file, then transcribed with another tool from this guide.
Watch out. A voice message holds someone else’s voice. Uploading it to an online transcription site passes their data to a third party without asking them.
On iPhone, Samsung and Mac
- iPhone. The Voice Memos app shows a transcript of recordings on recent models. Open the recording and tap the transcript icon.
- Samsung Galaxy. Voice Recorder offers Transcribe on models with Galaxy AI. The summary needs a Samsung account.
- Mac. Voice Memos also exists on Mac and picks up your iPhone recordings through iCloud. For a file from elsewhere, use Word or Whisper.
These transcripts mainly help you find a passage. For settings, length and limits, see Voice recorder on iPhone and Android.
Whisper: free, on your computer
Whisper is a speech-recognition model released as open source by OpenAI in 2022. It runs offline, on your machine, and handles many languages well. It is the only method that is free, quota-less and keeps the file at home.
The model comes in several sizes, from "tiny" to "large": the bigger, the more accurate and the slower. On a recent laptop, a mid-sized model is a good compromise. Without a strong graphics card, expect compute time in the order of the audio's length, sometimes more.
No need for the command line. Free apps wrap Whisper in an interface; noScribe, for example, was built for research interviews, separates speakers and comes with an editor to correct the text. Whisper also exports subtitles (SRT, VTT) for a video.
"Free" online services
Searching "transcribe audio to text free" returns dozens of sites. Their free tier is almost always capped in minutes, file length or number of exports. Those that work without sign-up have a catch: with no account, there is nobody to ask to delete the file. Read two things before uploading one:
- Where the service processes the audio. A server in Europe run by a US company is still subject to US law, as we explain in The CLOUD Act: why a server in France is not enough.
- What it does with your files. Some terms of use allow recordings to be used to improve models. Once audio has been used for training, there is no way back, as we show in GDPR cannot un-train a model.
For a public podcast, it does not matter. For a client meeting, an HR interview or a consultation, you are the one answerable for the data of the people recorded.
The AI voice recorder
An AI voice recorder records, sends the audio to a service that transcribes it, then returns a summary to your phone. It is the only method that removes every step in between: no file to transfer, no tool to open. It pays off if you record several meetings a week. Our guide How to choose a voice recorder compares the device families.
The Google Docs voice typing trap
Google Docs voice typing comes up often in advice. It transcribes what you dictate into the microphone, live, in Chrome. It does not read files. Playing a recording in front of the computer's microphone works poorly: the sound degrades, any notification interrupts the typing, and you must stay in front of the screen from start to finish.
What "95% accuracy" means
Tools like to display an accuracy rate. Translated into errors, it reads differently. Conversation runs at about 150 words a minute, so roughly 9,000 words in an hour of audio:
Even at 99%, an hour-long meeting leaves close to a hundred errors. Most are harmless. A few land on a name, an amount or a negation, and those are the ones you must find.
Get a reliable result, whatever the tool
- Bring the microphone closer. The distance between mouth and microphone matters more than the microphone's price.
- Avoid rooms that echo. An empty room with bare walls blurs consonants, and the engine picks the wrong words.
- One voice at a time. Overlapping speech is the first source of errors, for a machine and for a human.
- Pick the right language. A tool set to the wrong language "translates" instead of transcribing, or makes words up.
Proofreading without listening to everything again
- Search the text for the proper names you know: clients, colleagues, products. Each misspelt variant is fixed in one go with Find and Replace.
- Check every number against a document: quote, schedule, previous minutes.
- Reread negative sentences. A dropped "not" reverses a sentence without making it look wrong.
- Only replay the passages that sound off, at reduced speed.
Frequently asked questions
How do I transcribe audio to text for free?
With your phone's app if the recording came from it, with WhatsApp for a voice message, with Word if you have Microsoft 365, or with Whisper on your computer for everything else. Whisper is the only one with no quota and no upload.
Can I transcribe a WhatsApp voice message?
Yes, by turning on voice message transcripts in WhatsApp's settings. If your language is not offered, share the message as an audio file and transcribe it with another tool.
Can I transcribe audio in Word for free?
Transcribe is for Microsoft 365 subscribers, up to 300 minutes of uploaded audio per month. If your employer provides Microsoft 365, it costs you nothing extra.
Is there a tool with no sign-up?
Yes, several sites take a file without an account. Keep them for files with nothing at stake. For sensitive audio, Whisper on your computer needs neither an account nor an upload.
How do I get the text of a YouTube video?
If the video has captions, even auto-generated ones, open the description under the video and choose Show transcript. For your own videos, Whisper produces a subtitle file.
How long does automatic transcription take?
Online, a few minutes for an hour of audio. On your computer with Whisper, from a few minutes to over an hour, depending on the machine and model size. Always add proofreading time.
Key takeaways
- Free, yes: your phone, WhatsApp, Word if you have Microsoft 365, Whisper locally.
- The recording matters more than the tool.
- For professional audio, check where it is processed before uploading it.
- 99% accuracy still leaves about a hundred errors an hour: check names, numbers and negations.
Notus records at the press of a button and returns the transcript and summary to your phone in about thirty seconds, processed and hosted in France.
Discover Notus →