Free Audio to Text Transcription: What's Really Free and What's Not
Genuinely free transcription exists, but most tools called free are metered trials with caps buried under the headline. The real landscape splits three ways: free-with-limits plans, free features built into software you already own, and tools that are free because they run on your machine. Knowing which one you're looking at is the difference between a working workflow and a paywall in the middle of a project.
What Free Usually Turns Out to Mean
When a free plan promises hundreds of minutes, check whether that pool survives your first real work week. Otter gives you 300 minutes a month but caps every session at 30, so a long interview hits the wall before the recording ends. The minutes exist, and the shape of your work just doesn't fit inside them. That's the first trap: the pool is real, and the per-session cap is what actually limits you.
The second trap is the pool that looks generous until you read the per-file rule. Notta advertises 120 free minutes a month, but each conversation is capped at 3 minutes and the free tier has no export, which means chopping real audio into fragments and losing the context between them. The third trap is text you can't take with you: TurboScribe's free tier does about 3 transcripts a day up to 30 minutes each, and export is where free tiers most often stop. A transcript you can't move into a document is a demo, not a deliverable.
Read any free tier in that order: the monthly pool, the per-session cap, and whether the text leaves the app. Most "100% free" headlines survive none of those three questions.
The Built-In Options That Cost Nothing
Software you already own covers the live-speech direction for nothing. Google Docs Voice Typing and Windows Voice Typing (the Win+H shortcut) transcribe live speech with no subscription, which makes them the fastest path for dictation and quick notes. Their edges are hard ones, though: neither opens an audio file you already have, and neither separates speakers, so a two-person conversation comes out as one undifferentiated block.
Meeting platforms sit behind plan gates. Zoom's transcription is a cloud-recording feature, so the free plan records locally but produces no transcript, as covered in the Zoom transcription guide. Google Meet requires Workspace Business Standard or higher for its built-in transcription, per Google's plan structure. If your meetings live on the free tiers, the platform's own transcription isn't part of your option set.
There's a workaround for files with the live tools, and it's worth knowing even though it's clunky: play the recording out loud while the voice typing listens. It works for short clips, loses words whenever the room isn't quiet, and never sorts out who said what. Treat it as an emergency exit rather than a workflow.
The Unlimited Route: Running It Yourself
Whisper and whisper.cpp have no monthly quota, no per-file cap and no upload, because the model runs on your hardware and the audio never goes anywhere. That's the whole trick: you trade an evening of setup for a tool that never runs out of minutes. For anyone with hours of archived audio, or a daily habit of recording, it's the only category where "free" survives contact with real volume.
The middle path is Word's Transcribe if you already pay for Microsoft 365: 300 uploaded minutes a month, with unlimited uploaded minutes rolling out for Word on the web in commercial tenants. No command line, familiar interface, and the cap is the only thing standing between you and the local route.
The honest price of local is setup, not money. You'll install a tool, download a model, and spend an hour on your first run, and after that the marginal cost of transcribing an hour of audio is your electricity. If that trade sounds bad, the free tiers with their caps are still there, and you now know how to read them. For the full local workflow, see local AI transcription, and for teams that record meetings, the offline meeting note taker setup covers the same ground.
Matching the Tool to the Job
Live dictation goes to the built-in tools, because they're already open and they cost nothing. Short, occasional recordings fit the free tiers now that you know how to check the caps before committing. Daily volume, long interviews and archives belong to the local route, where the quota problem doesn't exist. The habit worth building is reading the caps before you commit a workflow to any of them, whichever category you pick.
Meetings are their own category, because the transcript needs speaker names to be useful. The platform's built-in transcription does that well when your plan includes it, and a local assistant like Shmeetings does it on macOS without the plan gate, capturing system audio and keeping the whole record on your machine. The wrong choice here isn't the tool you pick, it's the meeting that happened without anyone capturing it.
Interviews sit between the categories, and speaker labels decide where they land. A two-person interview transcribed without labels is a wall of text you have to re-read to attribute, which is exactly the work the transcript was supposed to remove. Since labeling is the feature free tiers most often hold back, an interviewer on a budget usually ends up local: run the file through Whisper for the words, or use a tool that handles diarization, and add the names while the conversation is still fresh in memory.
Frequently Asked Questions
Is there a truly free transcription tool?
Yes, in the local category: Whisper running on your own computer is free with no quota, no upload and no account. Everything else called free is a metered slice of a paid product. The local route trades setup time for unlimited use, which is the honest price.
How can I transcribe audio to text for free without limits?
Run Whisper or whisper.cpp locally. There are no minute caps because the compute is yours, and a modern laptop handles it fine, with a GPU making it faster rather than making it possible. It's the only route with no cap at all, because every other free option has a number attached somewhere.
Does Google have free transcription?
Google Docs Voice Typing is free for live dictation, but it can't process an existing audio file and doesn't separate speakers. Google Meet's built-in meeting transcription requires a paid Workspace Business Standard plan or higher. So Google's free transcription exists, but only for live speech into a Doc.
Is Otter really free?
The free plan is real but narrow: 300 minutes a month, 30 minutes per session, and 3 file uploads. Short meetings fit, and long interviews and heavy weeks don't. Whether it's "really free" depends on whether your work fits inside those three numbers.
What is the catch with free transcription tools?
The caps, in three forms: monthly minute pools that run out, per-session or per-file limits that make the pool unusable, and missing exports that lock your text inside the app. Speaker labeling is also commonly paywalled. Check those four things before you build a workflow on any free tier.