How to Transcribe Audio to Text: Every Method Compared (Free, Built-In and AI)
Dzianis Kukareka, Founder · Updated

To transcribe audio to text, give the recording to a tool that accepts files: Microsoft Word's Transcribe feature (with Microsoft 365), the recorder app on your phone, ChatGPT or Gemini, a local install of OpenAI Whisper, or an AI transcription service. Google Docs voice typing and Mac dictation only type what they hear through the microphone in real time, so they suit dictation, not recordings you already have. As a rule of thumb: a short file and a Microsoft 365 subscription points to Word, a recording made on your phone points to its built-in app, and a long recording with several speakers or in another language points to a transcription service or Whisper.
Below is every common way to transcribe audio, with the limit each one hits first. BlaBlaScribe, which publishes this guide, is one of the options, and we list its limits like everyone else's.
How to transcribe audio to text: all methods at a glance
| Method | Takes a file? | Where it runs | Main limit | Speaker labels | Cost |
|---|---|---|---|---|---|
| Google Docs voice typing | No, live microphone only | Chrome, Edge or Safari | Real time only | No | Free |
| Microsoft Word Transcribe | Yes: WAV, MP4, M4A, MP3 | Word for Microsoft 365 on Windows, Word for the web | 300 upload minutes a month | Yes (Speaker 1, Speaker 2) | Microsoft 365 subscription |
| iPhone or Mac Voice Memos | Its own recordings | On the device | Not in all countries, regions or languages | Not documented | Free |
| Mac dictation | No, live microphone only | On the Mac | Stops after 30 seconds of silence | No | Free |
| Pixel Recorder | Its own recordings | On the phone (re-transcribe may use Google servers) | 15 live languages on Pixel 6 and later | Pixel 6 and later | Free |
| Samsung Voice Recorder (Galaxy AI) | Its own recordings | Galaxy phone | Supported Galaxy models and languages | Not documented | Free on supported phones |
| ChatGPT | Yes, on paid plans | OpenAI cloud | 512 MB per audio file | Unreliable, per OpenAI | Paid ChatGPT plan |
| Gemini app | Yes | Google cloud | 10 minutes of audio; 3 hours with Google AI Pro or Ultra | Not documented | Free, more with AI Pro or Ultra |
| Whisper (local) | Yes | Your computer | Your hardware: about 1 to 10 GB of VRAM by model | No | Free (MIT license) |
| AI transcription service (e.g. BlaBlaScribe) | Yes | Vendor servers | Plan minutes; 8 GB per file in BlaBlaScribe | Yes | Free trial, then minutes or a plan |
| Human transcription service | Yes | Vendor | Turnaround in hours or days | Yes | Highest per minute |
Limits were checked on each vendor's help pages in November 2026. These pages change often, so open the linked page before you rely on a number.
If you are speaking right now, use dictation: Google Docs, your Mac or your phone keyboard.
If the recording is already on your phone and short, try the built-in recorder app first.
If the file is long, has several speakers or isn't in English, use a transcription service or Whisper.
Dictation vs transcription: which one do you need?
Dictation turns your voice into text as you speak. You talk into the microphone, words appear in the document, and you correct them as you go. It's built into Google Docs, macOS, iOS and Android keyboards, and it's free.
Transcription starts from a recording that already exists: an interview, a lecture, a meeting, a voice memo. A good transcript also carries timestamps and labels who said what, so you can find a quote and check it against the audio.
The common mistake is to play a recording through speakers next to a dictation tool. It works poorly: the tool hears room echo instead of a clean voice, you have to sit through the whole recording in real time, and you get one block of text with no speakers and no timestamps. If you have a file, use a tool that accepts files.
How to transcribe audio to text for free
You can transcribe audio for free with what's already on your devices, as long as you accept each tool's catch.
| Free option | The catch |
|---|---|
| Voice Memos (iPhone, Mac) | Works on recordings in the app; not available in every country, region or language |
| Pixel Recorder, Samsung Voice Recorder | Works on recordings made in the app; only on supported phones |
| Gemini app without a subscription | Up to 10 minutes of audio in total per prompt |
| Whisper on your computer | Needs Python and ffmpeg, a capable machine and no speaker labels |
| Free tiers of online services | A capped number of minutes |
Two things that look free often aren't. Word's Transcribe feature is part of a Microsoft 365 subscription, and ChatGPT only accepts audio files on paid plans. For a rundown of free online services and where each one stops, see the best free AI transcription tools and their real limits.
BlaBlaScribe gives 30 free minutes once when you sign up: short files are transcribed in full, and for long ones you see the first 10 minutes.
How to transcribe audio to text in Word
Microsoft Word can transcribe an uploaded recording with its Transcribe feature. According to Microsoft's help page, it's available in Word for Microsoft 365 on Windows (commercial tenants) and in Word for the web, where it only works in Microsoft Edge and Chrome and needs an internet connection.
Open a document and go to Home → the Dictate menu → Transcribe.
Choose Upload audio and pick a WAV, MP4, M4A or MP3 file.
Keep the Transcribe pane open while it works (Microsoft says it can take about as long as the audio), rename speakers if needed, then select Add all to document.
The limits: Microsoft 365 subscribers get 300 minutes of uploaded audio per month (30,000 with a Microsoft Copilot license), and Transcribe supports more than 80 locales. Speakers are separated and labeled Speaker 1, Speaker 2 and so on, and you can rename a label everywhere at once. Microsoft doesn't state a maximum file size.
Can you transcribe audio in Word on Mac?
Microsoft's help page documents Transcribe for Word on Windows and Word for the web, and doesn't mention Word for Mac. On a Mac, the documented route is Word for the web in Chrome or Edge.
How to transcribe audio to text in Google Docs
Google Docs has voice typing, not file transcription. Open a document, go to Tools → Voice typing and click the microphone. It works in current versions of Chrome, Edge and Safari, and Google's language list has about 120 entries including regional variants. Voice commands such as "new paragraph" work only in English, and a Workspace admin can turn the feature off (Google Docs Editors Help).
Google's help page describes microphone input only, with no option to upload an audio file. To get a recording into Docs, transcribe it somewhere else and paste the text, or export a DOCX and open it in Docs.
How to transcribe audio to text on iPhone
Voice Memos
Voice Memos can show a transcript of a recording. Open the recording, view its transcript, then search it or select and copy the text into another app. Apple notes that audio transcription isn't available in all countries or regions (Apple's Voice Memos guide), and the feature depends on your device and language, so check Apple's iPhone User Guide for your model.
Dictation on the keyboard
The microphone key on the iPhone keyboard is dictation: it types what you say live. It doesn't transcribe files.
For long recordings, or files that didn't start in Voice Memos (a voice note you saved from a chat app, an M4A someone emailed you), upload the file to a service instead. A voice memo to text tool takes the M4A as it is.
How to transcribe audio to text on Mac
Mac dictation is switched on in System Settings → Keyboard → Dictation. You start it with the microphone key, your dictation shortcut or Edit → Start Dictation. Apple says you can dictate text of any length without a timeout, but dictation stops after 30 seconds without speech, and it isn't available in every language or region (Apple's dictation guide). It doesn't open audio files.
For files on a Mac, Voice Memos shows a transcript of recordings in the app (with the same regional caveat), and anything else goes to Word for the web, Whisper or a transcription service.
How to transcribe audio to text on Android (Pixel and Samsung)
Google Pixel: Recorder
The Recorder app on Pixel transcribes as you record. According to Google's Pixel help, live transcription covers 8 languages on Pixel 3 to Pixel 5a and 15 on Pixel 6 and later. You can re-transcribe a recording in about 40 more languages, but Google warns that re-transcribing may process the audio on Google servers. Speaker labels need a Pixel 6 or later, and summaries a Pixel 8 or later. The help page doesn't describe importing audio files from elsewhere.
Samsung Galaxy: Voice Recorder with Galaxy AI
On Galaxy phones with Galaxy AI, open a recording in Voice Recorder, tap the Transcript assist icon at the top, then tap Transcribe (Samsung support). Extra languages are downloaded from the same menu, and the list depends on your region and software version.
Other Android phones
Recorder apps on other brands vary, and the keyboard microphone is dictation, not file transcription. For a saved file, use a transcription service, Gemini or Whisper on a computer.
Can ChatGPT or Gemini transcribe audio to text?
Yes, both can, with limits.
ChatGPT. OpenAI's article on uploading files and audio to ChatGPT says audio uploads are available with paid subscriptions and workspaces, not on the Free plan. Audio files can be up to 512 MB in WAV, MP3, OGG, audio-only WebM, PCM, FLAC, AAC, M4A or audio-only MP4. Files identified as video are rejected, very long recordings may time out, and OpenAI warns that transcripts may contain errors and speaker identification may be unreliable.
Gemini. Google's Gemini Apps help says total audio length can be up to 10 minutes, or up to 3 hours with Google AI Pro or Ultra, and you can upload up to 10 files in one prompt.
Both return the transcript as a chat reply. Timestamps and speaker labels depend on how you prompt and aren't guaranteed, and on long files a language model can skip or paraphrase passages. That's fine for getting the gist of a voice note, and risky for quotes, subtitles or anything you'll publish. We haven't run our own accuracy tests for this article, so we're not giving percentages.
How to transcribe audio to text locally with Whisper
OpenAI Whisper is an open-source speech recognition model you can run on your own computer, so the audio never leaves it. The Whisper repository releases the code and model weights under the MIT license. You need Python and ffmpeg, and a graphics card helps a lot.
| Model | Required VRAM | Relative speed |
|---|---|---|
| tiny | ~1 GB | ~10x |
| base | ~1 GB | ~7x |
| small | ~2 GB | ~4x |
| medium | ~5 GB | ~2x |
| turbo | ~6 GB | ~8x |
| large | ~10 GB | 1x |
Figures are from the Whisper README, measured on English speech. The turbo model is the default and isn't trained for translation. OpenAI notes that Whisper's performance varies widely by language, and the README doesn't include speaker labels, so you get text and timestamps but not who said what.
If you're a developer, OpenAI's hosted speech-to-text API takes files up to 25 MB per request. It lists whisper-1 and the gpt-4o transcribe models as deprecated with a shutdown date of February 26, 2027, and recommends gpt-transcribe. For how local Whisper compares with paid tools in practice, see Whisper vs paid transcription tools.
What is the best way to transcribe audio to text?
For recordings longer than half an hour, with several people talking or in a language other than English, built-in tools run into their limits: minutes per month, a short list of languages, no speaker labels, or files they won't open. That's when a transcription service or Whisper is the better route.
What to check in a transcription service
Format and size. Does it take your file as it is, or do you have to convert it? See what audio format is best for transcription.
Your language. Check the language list, not just the count.
Speaker labels. Essential for interviews and meetings.
Timestamps and exports. TXT or DOCX for documents, SRT or VTT for subtitles.
Privacy. Where the audio is processed, how long it's kept and whether it's used to train AI.
Price per hour of audio, not per month.
Audio quality matters more than the tool, which is why AI transcription accuracy varies so much between files. For a comparison of paid services, see the best AI transcription software.
In BlaBlaScribe you upload a file of up to 8 GB, choose from 120+ languages and get a transcript with speaker labels and word-level timestamps, which you can export as TXT, DOCX, SRT or VTT. Files are processed on servers in Germany and aren't used to train AI. The limits: it doesn't accept FLAC or AVI, so convert those first; the free account exports SRT and TXT only; and full transcripts of long files need a minute pack (from $5 for 50 minutes) or a plan from $14 a month (pricing). If that fits, the audio to text tool is where you upload.
How to transcribe audio with multiple languages
Most tools expect one language per file, and you pick it before you start. If speakers switch languages mid-sentence, quality drops in almost any tool. If each person speaks one language, it often works better to cut the recording into single-language parts and transcribe each with the right language set, or to check whether your tool has a mixed-language mode and test it on a short clip first.
How to transcribe audio to text from a video or YouTube link
A video file is just audio with pictures, and most file-based tools accept MP4: Word Transcribe, Gemini and transcription services all do. ChatGPT is the exception, since its audio uploader rejects files identified as video. If you only need the audio, a free MP4 to MP3 converter extracts it in your browser. For videos already on YouTube, see how to get the transcript of any YouTube video.
How to transcribe audio to text manually
Manual transcription still makes sense for very short clips, for audio no tool handles well, or when you're checking an automatic draft line by line. The W3C Web Accessibility Initiative describes common practice:
Write down what is said without correcting grammar or adding words.
Use a speaker's full name the first time, then one name after that.
Describe relevant non-speech sounds in parentheses, such as (laughter) or (door slams).
Mark speech you can't make out as [unintelligible].
A simple layout that follows those rules:
[00:00:00] Maria Lopez: Thanks for coming in. Can you start with your role?
[00:00:06] James Carter: Sure. I run the support team (laughs), all four of us.
[00:00:14] Maria: And how long have you been there?
[00:00:16] James: Since [unintelligible], so about six years.
Slow the playback down, use a pause shortcut so your hands stay on the keyboard, and add a timestamp every 30 to 60 seconds so you can find passages later. Expect several hours of work per hour of audio, more with crosstalk. W3C also points out that automatic transcription is a reasonable starting point, but you need to correct it: missing a single word like "not" can reverse the meaning.
Which method should you use?
You're dictating now: Google Docs voice typing, Mac dictation or your phone keyboard.
The recording is on your phone, short and in one language: Voice Memos, Pixel Recorder or Samsung Voice Recorder.
You have Microsoft 365 and fewer than 300 minutes a month: Word Transcribe.
The audio can't go to the cloud and you have a capable computer: Whisper locally.
Long recording, several speakers, or you need subtitles or translation: a transcription service.
Legal, medical or broadcast text where a mistake is costly: an automatic draft plus a careful human review, or a human service.
For step-by-step workflows, see how to transcribe an interview and how to transcribe a lecture.
Frequently asked questions
What is the easiest way to transcribe audio to text?
Give the file to a tool that accepts recordings: Word Transcribe if you have Microsoft 365, your phone's recorder app if you recorded there, or an AI transcription service. Dictation tools like Google Docs voice typing need you to speak live.
Can I transcribe audio to text for free?
Yes. Voice Memos, Pixel Recorder and Samsung Voice Recorder transcribe their own recordings, Whisper runs free on your computer, and Gemini takes up to 10 minutes of audio without a subscription. Online services usually offer a limited number of free minutes.
Can Google Docs transcribe an audio file?
Not directly. Voice typing listens to the microphone in real time, and Google's help page describes no file upload. Transcribe the file elsewhere and paste or import the text.
Can ChatGPT transcribe audio?
Yes, on paid plans: you can attach an audio file of up to 512 MB, while the Free plan can't upload audio. OpenAI warns that transcripts may contain errors and speaker identification may be unreliable.
How do I transcribe a recording with several speakers?
Use a tool with speaker labels, also called speaker diarization: Word Transcribe, Pixel Recorder on a Pixel 6 or later, or a transcription service. Then check the labels at the points where people interrupt each other.
Is it safe to upload recordings to a transcription service?
Check where files are processed, how long they're kept and whether they're used to train AI before you upload. For very sensitive audio, on-device apps or Whisper on your own computer keep the recording off third-party servers.
Turn your recordings into text you can use
Upload audio or video, paste a YouTube link, or record in the browser — and get a speaker-labeled transcript in minutes, in 120+ languages.
Try BlaBlaScribe for freeNo credit card required · Stored in the EU · Never used for AI training





