Whisper, Otter, Granola, ElevenLabs: which free tool to transcribe audio to text in French, with the real limit of each plan and its price.
Audio-to-text transcription today reaches professional quality in French, as long as you choose the tool based on the source. For a long or confidential file, Whisper locally remains unbeatable: free, no duration limit, nothing leaves your computer. For video calls, Granola or Otter do the job without paying, as long as you stay under their caps. When free is no longer enough, the first useful tier costs between 6 and 15 dollars per month depending on the tool.
Every free transcription tool lets you transcribe, but it recovers its margin on another axis: duration, volume, language, or export.
Otter cuts meetings to 30 minutes on its free plan. Read AI stops at 5 meetings per month. Wispr Flow counts words by the week. ElevenLabs gives you credits but forbids commercial use. Google Docs dictation reads no file, only your mic in real time.
The classic trap: test a tool on a 10-minute audio, be blown away, then discover the limit on your first real one-hour interview. The limit is always on the pricing page, rarely on the homepage.
The right question is not "which tool transcribes best," but "which limit bothers me least." Someone transcribing two podcasts a month doesn't have the same needs as a team recording ten meetings a week.
I judged these tools on recordings typical of a small business: video calls in French, multiple speakers, industry jargon, and proper names.
I don't give error rates to the word, which would be false precision that depends too much on the microphone and accent. I looked at what matters day to day: passages to correct by hand, handling overlapping voices, speaker identification, and wait time before getting the text.
The finding is surprising: on clean audio, quality gaps between tools are small. All rely on speech recognition models of the same generation. Errors concentrate on brand names, acronyms, and moments when two people talk at once.
What really separates tools is everything else: duration limit, ability to import a file, local or cloud processing, and quality of the summary produced after transcription.
For transcribing an already-recorded file, Whisper locally and ElevenLabs Scribe are the two most solid options without paying.
It's my default solution. I transcribe my voice notes and interviews with Whisper on my Mac, in its version optimized for Apple chips: an hour of audio comes out in a few minutes, never leaving the machine. Code and models are documented on the official Whisper repository.
For a Zoom, Meet, or Teams meeting, Granola and Otter offer the best ratio between quality and generosity of the free plan.
A French option also exists: Meet by IAxion, which I tested in detail in my Meet by IAxion review, with data hosting in France.
Dictation transcribes your voice in real time, which makes it a writing tool rather than a tool for transcribing existing recordings.
This table summarizes, tool by tool, the free plan limit that will block you first and the price of the first tier.
| Tool | Audio Source | Main Free Limit | Processing | First Paid Tier |
|---|---|---|---|---|
| Whisper (local) | File | None | Local | Free |
| ElevenLabs Scribe | File | 10,000 credits/month, non-commercial | Cloud | $6/month |
| Meetily | Meeting | Technical setup | Local | $10/month (cloud) |
| Granola | Meeting | Capped history | Cloud | $14/month |
| Otter.ai | Meeting, file | 30 min per meeting, 3 imports lifetime | Cloud | $8.33/month (annual) |
| Read AI | Meeting | 5 meetings/month | Cloud | $15/month (annual) |
| Wispr Flow | Dictation | ~2,000 words/week | Cloud | $15/month |
| Google Docs Dictation | Dictation | Mic in real time only | Cloud | Free |
The right choice depends on the nature of your audio: recorded files, video meetings, or text to dictate.
You transcribe interviews, podcasts, or lectures: install Whisper locally. It's free forever, and your confidential files stay with you. If you refuse any installation, ElevenLabs Scribe does the job for personal use.
You do meeting after meeting: Granola if you want to stay discreet and keep control of your notes, Otter if you prefer a fully automatic summary. For meetings longer than 30 minutes, Otter will quickly push you toward paid.
You work in a company that cares about its data: Meetily locally, or a solution hosted in France. A recorded voice is personal data under GDPR, presented by the French data protection authority: participants must be informed.
You want to write faster: Wispr Flow or Google Docs dictation. These aren't file transcription tools, but they beat the keyboard hands down.
For more on free AI assistants, I've listed truly usable free AI applications, and the guide to AI tools for founders covers the rest of the toolkit.
Going paid is justified the moment the free limit costs you more time than the subscription.
Three signals don't lie. You're splitting your files to work around a duration limit. You're juggling two free accounts. You're manually retyping a summary because export is restricted. Each of these workarounds costs more than 10 dollars of your time per month.
The cheapest tier remains ElevenLabs Starter at 6 dollars per month, useful for commercial file use. For meetings, Otter Pro annually is the most affordable. Granola and Read AI fall between 14 and 15 dollars, a fair price if you do multiple video calls a day.
Before paying, check one thing: an assistant like ChatGPT or Gemini already accepts audio files and can pull a summary from them, which sometimes suffices for a one-off need.