You can transcribe a 60-minute recording for $0. The catch isn't really a catch. It's a stack of small limits that add up: a length cap, a speaker cap, the wrong export format, a slow turnaround, or a privacy policy that says "we may use your audio to improve our models."
The free tier is real. It's also not the same product as the paid one. Knowing what you actually get keeps you from wasting an afternoon on a tool that won't finish your file.
What "free" actually means for AI transcription
Three different things hide under the word "free," and they behave nothing alike.
- Free forever — a real free tier with a monthly cap (e.g., Otter's 300 minutes/month, 30 minutes per recording).
- Free trial — full features for 7 or 14 days, then a paywall drops.
- Free and open-source — you run the model yourself (Whisper, faster-whisper). $0 in software cost, but your laptop and your time are the resources.
The first two are SaaS marketing. The third is engineering. The limits look completely different depending on which one you're on.
How accurate is free AI transcription?
Clean audio, one speaker, a quiet room: 90–95% word accuracy is realistic across the major free tools. Good enough that you'll skim and lightly edit, not retype from scratch.
The number falls fast when reality intrudes. Two speakers talking over each other on a phone call? 80–85%. A noisy cafe interview? Lower. A heavy accent or technical jargon? Lower still. Free tools usually run the same underlying models the paid ones do — accuracy isn't where they cut corners. For the honest numbers, see transcription accuracy: what to expect.
What are the typical limits on a free plan?
A pattern shows up across most free SaaS tiers:
- Monthly minute cap — usually 300–600 minutes per month.
- Per-file length cap — often 30–40 minutes per recording. A long lecture or podcast hits this first.
- Limited exports —
.txtonly, no.srt, no.docx, no timestamps. - No speaker labels, or labels limited to two speakers.
- No translation, or translation locked to one or two languages.
- Slower queue priority — your file waits behind paying customers.
- A "your data trains our model" clause — buried in the privacy policy.
Each tool picks 3–5 of these to enforce. The ones that hurt depend on what you're actually transcribing.
What features get locked behind the paywall?
The paid tiers gate the features that turn a transcript into a useful artifact:
- Speaker diarization beyond two speakers. If the term is new, speaker diarization in plain English covers it.
- Word-level timestamps for video editing or subtitle work.
- Subtitle formats —
.srt,.vtt, burned-in captions. - Bulk upload — drop 10 files at once instead of one.
- Custom vocabulary — names, brands, technical terms.
- API access — programmatic transcription inside a workflow.
- Team sharing and admin controls.
One short file, one speaker, plain text out — free works. Anything past that and the paywall starts to bite.
Is free AI transcription safe for sensitive recordings?
Read the privacy policy before you upload. Free tools often retain audio to improve their models. Some let you opt out, some don't. Some are clear, some aren't.
Things to watch for:
- "We may use de-identified content to improve our services" — assume your audio is in the training pool.
- "Data retained for X days" with no opt-out — fine for a podcast, not fine for a deposition.
- No mention of data location or transit encryption — pass.
For anything regulated (legal, medical, HR investigations), assume free isn't the right tool. Run a local Whisper model on your own machine, or pay a vendor that will sign a BAA. The same applies to journalist interviews with sources who were promised confidentiality.
When should you upgrade to a paid tool?
A short test: if any of these is true more than once a month, you'll save time and money by paying.
- You're hitting the 30-minute per-file cap and splitting recordings.
- You need speaker labels for 3+ people.
- You need
.srtfiles or timestamps for a video edit. - You upload more than 10 hours a month total.
- You're transcribing anything you wouldn't want sitting on a vendor's server.
Paid AI transcription runs roughly $0.10–$0.50/min on most platforms. A monthly subscription only pays off if you're doing real volume. We broke this down in how much AI transcription actually costs and looked at the subscription trap in is there a cheaper Otter alternative.
Paste any public link or upload a file and get a clean transcript in minutes. First 3 clips every month are on us — no card required.
What free tool fits which use case?
A practical map, not a ranking:
- Short clean recording, one speaker, plain text out — Google Recorder (Android, on-device), Apple's built-in Voice Memos transcription (iOS/macOS).
- A meeting with auto-join — Otter's free tier, or Microsoft Teams' built-in transcription if your org has it enabled.
- Anything you don't want a vendor to keep — Whisper or faster-whisper on your own laptop. Zero cost in dollars, an evening of setup the first time, worth it if privacy or volume matters.
- A quick one-off where you don't want an account — transcribe a video without signing up; you pay per minute for what you actually upload, no subscription, no data retention beyond your session.
Pick the tool that matches the bottleneck you're actually hitting — length, speakers, exports, privacy, or volume. Free is plenty for the first one. The rest, eventually, aren't.
