Transcribe Audio to Text Free: 5 Methods That Actually Work

Transcription used to be a paid service you sent a file to and waited a day for. Now the same job runs on your laptop, in your browser, or inside software you already pay for. The catch is that the free options all have a shape: one does live speech only, one caps you at a few hundred minutes, one needs you to be comfortable with a command line.

Pick by what you actually have: a meeting that's about to start, a recording on your phone, or ten hours of interviews.

Transcribe audio to text free: match the method to the job

You have	Use	Free limit
A live conversation or dictation	Google Docs voice typing	Unlimited, live speech only
A recorded file and Microsoft 365	Word Transcribe	300 minutes of uploads a month
Meetings to transcribe	Otter.ai free	300 minutes a month, 30 per conversation
Lots of files, any language	Whisper on your computer	Free, open source
A video you'll publish	YouTube auto-captions	Free

1. Google Docs voice typing: free and unlimited, but live

Open a Google Doc, click Tools → Voice typing, and talk. Google's help page says it works in the latest versions of Chrome, Edge and Safari, and supports over 100 languages and regional accents, though voice commands themselves are English-only.

The limitation to understand: it transcribes what your microphone hears right now. It isn't built to transcribe a file you already have. People do get around this by playing a recording through their speakers, but the quality drops because you're transcribing a recording of a recording. For dictating notes, a first draft or a letter, it's free and unlimited.

2. Word's Transcribe: 300 minutes a month with Microsoft 365

If you pay for Microsoft 365, you already have a file transcription service. Microsoft's Transcribe documentation says subscribers "can transcribe a maximum of 300 minutes of uploaded audio per month," with a Copilot licence raising that to 30,000 minutes. It accepts .wav, .mp4, .m4a and .mp3, supports 80+ locales, and runs in Word for the web and on Windows, with some desktop restrictions.

Five hours a month covers most people's interviews or lecture recordings, and the transcript lands in a document with speakers separated and timestamps you can click.

3. Otter.ai: for meetings

Otter.ai's free Basic plan gives 300 transcription minutes a month with a 30-minute cap per conversation, according to its pricing page. It joins Zoom, Teams and Google Meet on paid plans and gives you a searchable transcript with a summary.

If the audio is meetings rather than files, start here, and Fireflies is the main alternative: its free plan allows unlimited transcription with limited AI summaries. Both appear in our AI productivity tools ranking.

4. Whisper: free, unlimited, and it never leaves your laptop

OpenAI's Whisper is a speech recognition model released under the MIT licence that runs locally. Install it with pip install -U openai-whisper and transcribe as many files as you like, in dozens of languages, with no monthly cap and nothing uploaded anywhere.

The trade-offs are honest ones: you need Python and a terminal, and speed depends on your hardware. The project's own table ranges from the tiny model at roughly 1 GB of VRAM to the large model at about 10 GB, with a turbo model in between. On a laptop without a dedicated graphics card, the small or base model is the realistic choice.

This is the only option on the list where confidential audio never leaves your machine, which makes it the right answer for medical, legal or HR recordings.

5. YouTube auto-captions: the free shortcut

If the audio belongs to a video you're publishing anyway, upload it to YouTube as unlisted and let YouTube generate captions, then download the transcript. It's free, it handles long files, and the accuracy is reasonable for clear speech. Don't use it for anything private: the file goes to YouTube.

Getting a transcript worth editing

Every tool on this list does better with better input:

1. Record close to the speaker. Distance costs more accuracy than any setting.
2. Avoid overlapping speech. No model handles two people talking at once well.
3. Say names and numbers clearly, and fix them first when editing; those are what people search for later.
4. Clean the audio first if it's noisy. Our guide to removing background noise covers the free options.
5. Budget editing time. A good transcript still needs a pass: punctuation, speaker labels, technical words.

Indian English, regional accents and code-switching between English and Hindi all reduce accuracy in practice. Test two tools on the same five-minute clip before committing to a long project.

If you need to edit audio too

Descript transcribes and lets you edit the audio by editing the text, which is the fastest workflow for podcasts. Its free plan is limited (60 media minutes a month, watermarked exports), so it suits trying the workflow rather than production work. See our AI voice and audio tools ranking for more.

FAQ

What is the best free way to transcribe audio to text?

For a recorded file, Word's Transcribe if you have Microsoft 365 (300 minutes a month), or Whisper on your own computer for unlimited free transcription. For live speech, Google Docs voice typing is free and unlimited.

Can Google Docs transcribe an audio file?

Not directly. Voice typing transcribes live speech through your microphone. Playing a recording into it works poorly; use Word Transcribe, Otter or Whisper for files.

Is Whisper really free?

Yes. OpenAI released Whisper under the MIT licence and it runs locally, so there's no subscription and no upload. You need Python, and bigger models need more memory.

How accurate is free AI transcription?

Good on clear, single-speaker audio and noticeably worse with background noise, overlapping speakers or heavy accents. Expect to edit names, numbers and technical terms whichever tool you use.

How do I transcribe a Zoom or Google Meet call for free?

Otter.ai's free plan covers 300 minutes a month, and Fireflies' free plan allows unlimited transcription with limited AI summaries. Tell participants you're recording before you start.