Transcribe audio to text with AI: tools and steps

In brief

To turn an audio file into text, drop it into a transcription app: MacWhisper on a Mac does it for free without leaving your computer; Otter, Noota, or tl;dv do it online, with a free monthly limit. Pick the language yourself, proofread names and numbers, and keep sensitive recordings on your own device.

Which tool for which recording?

It depends on where the sound comes from: a file you already have (an interview, a voice memo, a lecture) or a meeting happening now. Prices and limits come from our tool pages, checked on October 19, 2026 (October 3 for tl;dv).

Tool	What it’s for	Free	Good to know
MacWhisper	An audio or video file, on a Mac	Transcription of your files, with no subscription, in about a hundred languages	Everything happens on your Mac. No Windows version. Pro: €64, one-time payment
Superwhisper	Dictation, and file transcription with Pro	Dictation in all your apps, unlimited on-device models	File transcription needs Pro: $8.49 per month
Otter	Online meetings, imported files	300 minutes per month, 30 minutes at most per conversation, 3 imported files in total	You pick French in the settings yourself; Otter doesn’t detect the language
Noota	Meetings and calls, in French	300 minutes per month, meetings kept for one month	Data hosted in Europe; more than 50 languages
tl;dv	Zoom, Meet, and Teams meetings	Recordings and transcripts with no limit on number, kept for 3 months	You have to say which language is actually spoken

For a single file on a Mac, MacWhisper is enough. For meetings that come back every week, see our guide to AI for meetings.

Transcribing an audio file, step by step

Gather the file. An mp3, m4a, or wav, or a video: MacWhisper also takes mov and mp4 files, and YouTube links.
Pick the language. Otter and tl;dv don’t guess it: a meeting that switches from French to English will be transcribed badly.
Start the transcription. In MacWhisper, you drag the file into the window and the text appears. In an online tool, you import the file or let the assistant join the meeting.
Proofread against the sound. Proper names, acronyms, and numbers are the first to go wrong. Speaker detection also gets it wrong sometimes.
Export. Free MacWhisper gives .txt and subtitles; Pro adds .docx, .pdf, .md, and .html. At Otter, exporting to txt, pdf, docx, and srt needs the Pro plan.

A transcript is raw text, not meeting minutes. To get a summary, paste it into your assistant with a clear instruction:

“Here is the transcript of an interview. Write a one-page summary: the decisions, the open points, and the actions with the name of the person responsible. If something is missing, write ‘to be confirmed’ instead of guessing.”

Does your recording leave your computer?

An interview can contain client names, employee data, or a trade secret. Depending on the tool, the sound stays with you or travels.

On your device: MacWhisper, according to its maker, sends no audio file or text anywhere else. Superwhisper does the same with a “Local” engine.
In Europe: Noota hosts data in Europe and says it doesn’t sell or share it; tl;dv, a German company, stores recordings in Europe and says it doesn’t use them to train AI.
In the United States: Otter stores your recordings in the United States and uses them to train its own AI, after removing what identifies you.

For sensitive content, choose the on-device tool. Our pillar page AI and GDPR at work sets the framework, and the guide what not to give an AI at work details what to keep to yourself. Also remember to tell the people you record.

The limits to know

Languages. Otter only understands six languages: English, Spanish, French, German, Japanese, and Simplified Chinese. Its interface is in English.
Free caps. Otter cuts off at 30 minutes per conversation and keeps only 25 conversations readable. Noota keeps meetings for just one month, tl;dv for three: download what you want to keep.
Mac only. MacWhisper doesn’t exist on Windows. On a PC, use an online tool.
Proofreading. No tool replaces a quick listen to the passages that matter.

What Nova brings

Nova is a subscription software, in French, that shows what these tools do in your job: journalist, researcher, recruiter, or legal assistant. Each tool page lists the plans, the limits, and the date they were checked, with demos to redo step by step, instruction ready. For the whole team: AI training for employees. See the interactive demo.

Frequently asked questions

Can you transcribe audio to text for free?

Yes. MacWhisper transcribes your files with no subscription on a Mac, in about a hundred languages. Online, Otter and Noota offer 300 minutes per month, and tl;dv offers transcripts with no limit on number, within each plan’s retention limits.

Is an AI transcript reliable?

It is good on substance, less so on proper names, acronyms, and numbers. Proofread those passages while listening to the recording before you share anything.

How do you transcribe audio on Windows?

MacWhisper only exists on Mac. On a PC, import the file into an online tool like Otter, picking the language yourself, or run Noota or tl;dv on your meetings.

Does my audio file leave my computer?

Not with MacWhisper, according to its maker, nor with Superwhisper on a “Local” engine. With Otter, Noota, or tl;dv, it goes to their servers: in the United States for Otter, in Europe for Noota and tl;dv.

What is the difference between a transcript and meeting minutes?

A transcript reproduces everything that was said, word for word. Minutes keep the decisions and the actions. Transcribe first, then ask your assistant for the summary, with a precise instruction.

Sources

MacWhisper, Publisher
MacWhisper privacy policy, Publisher
Otter pricing, Otter.ai
Noota pricing, Noota
tl;dv pricing, tl;dv
Superwhisper, Publisher

Transcribe audio to text with AI, step by step