Your voice notes stay your voice notes.
AI analysis with no data storage. Full security and privacy for your messages.
Free to try. No download, no sign-up.
Abrège
online
Forwarded message
- Wants to move Sunday brunch to Saturday noon
- Her cousin Marie is coming with her new boyfriend
- Needs your confirmation before 8pm tonight
What we keep, and what we don't
The question isn't what we
transcribe
.
It's what we keep afterwards.
The audio file is deleted from our servers right after the summary is sent, with no copy kept. Transcription goes through OpenAI's API, the only step handled outside our servers, and that's detailed further down this page.
Summary · 6 s
- • Meeting moved to Tuesday 2pm.
- • Confirm with the client before tonight.
- • Bring the file we prepared.
- • Send the signed contract if ready.
Privacy isn't a marketing promise
Many "free" transcription services pay their way by using your data to train their models. Your personal voice notes end up in datasets, accessible to employees, sometimes leaked.
Abrège makes the opposite commitment: the audio file is deleted immediately after transcription, with no copy kept, and our servers are hosted in Europe. Transcription is done via OpenAI's API, which states it doesn't train its models on data sent via its API.
In practice: your voice note reaches our servers, transcription is done via OpenAI's API, then the summary comes back to you and the audio is erased from our servers. We keep no copy.
How it works
Three seconds, zero effort.
Forward the voice note
To Abrège, on WhatsApp. Yours or a friend's.
It listens for you
Six seconds, then the audio is deleted.
You read what matters
The key points, the thing to do. Nothing else.
The concrete commitment
« A lawyer uses Abrège to transcribe her clients' voice notes. Confidentiality is a professional duty. She asked us for a technical audit before subscribing. We provided the documentation: no audio logs, deletion within 1 minute, Hetzner servers in Europe, TLS-encrypted transfers and transcription via OpenAI's API (which doesn't train its models on API data). She's been a client ever since. »
Amnesiac by design.
We keep nothing.
The moment your summary goes out, the audio file is destroyed from our servers. Unrecoverable, no copy kept. We don't want your data, just to save you time.
One thing to be clear about: to be transcribed, the audio passes through OpenAI's API (Whisper). That's the only step where it leaves us. See the details .
Deleted immediately
The audio is erased right after the summary is sent.
Never in your groups
Abrège only sees the voice notes you forward it. Never your other chats.
Encrypted in transit
Everything travels over TLS 1.3, impossible to intercept on the way.
Thousands of hours handed back to real life.
The time our users no longer spend listening to rambling monologues. Added up, it starts to count.
Format acceptés :
Frequently asked questions
About Your voice notes stay your voice notes.
- How long do you keep the voice notes?
- Strictly as long as processing needs, generally under a minute. No copy is kept after that.
- Do you use my voice notes to train your AI?
- We don't. Transcription goes through OpenAI's API, which for its part states it doesn't train its models on data sent via its API. And the audio file is deleted right after processing, with no copy kept.
- Are you GDPR compliant?
- Yes. Our servers are hosted in Europe. Transcription is done via OpenAI's API (United States), governed by its standard contractual clauses, and Abrège stores no interaction, message, or voice note.
- And is the transcription itself stored?
- The text result is sent to you via WhatsApp. On the Abrège side, we keep a short history for debugging and support, with no human access by default.
Pricing
Free to try. 3€ to never go without.
Start free, step up whenever you need to.
Free
Perfect for putting the AI to the test on your first voice notes.
0€
- 5 free summaries / month
- Limited settings
Starter subscription
🎁 2 months free
⚡️ Price locked for life
- 50 summaries / month
- Full Dashboard access
- Full customisation
Need more? Plans for heavy users (up to 500 summaries / month) are available. See all plans →
Cancel in one click from your account.
FAQ
Questions? We've got answers.
- No, never. Abrège works like any regular WhatsApp contact: it only sees the messages you forward or send it directly. It has no access to your other conversations, your groups, or your address book. WhatsApp itself guarantees this: a third-party app cannot read chats it hasn't been invited to. When you forward a voice note, only that one file reaches Abrège, nothing else. And once the summary is sent back, the audio file is immediately deleted from our servers, with no copy kept. To be transparent: to be transcribed, that audio passes through OpenAI's API (Whisper), the only step where it is handled outside our servers. The rest of your messages stays strictly between you and the people you talk to. In short: Abrège can only read what you explicitly choose to show it, and nothing more.
- When you forward a voice note to Abrège, it goes through three quick steps. First transcription: the audio is turned into text through OpenAI's API (Whisper), today one of the most accurate models around. Then the summary: a language model pulls out the key points and the action to remember. Finally delivery: you get the result right in your chat, in a few seconds. On the data side, the principle is simple: as soon as the summary is sent to you, the original audio file is deleted from our servers, with no copy kept. The text transcription stays available in your dashboard for as long as your account exists, and you can delete it whenever you like. Our servers are hosted in Europe and everything is encrypted in transit (TLS 1.3).
- Yes, rather well. Abrège relies on Whisper, OpenAI's transcription model, trained on a huge amount of real spoken language, not just textbook speech. In practice it handles regional and non-native accents very well. Everyday language comes through too: slang, casual expressions, hesitations (“er…”, “um…”), broken sentences or fast delivery. That's exactly the point of a genuine, recent AI model compared with basic word-for-word transcription. The limits: on very noisy audio, with several people talking at once, or very rare proper nouns, the transcription may make a few approximations. Even then, you still get your summary, and you can ask for the full text to check the details.
- It's commitment-free, and built to stay simple. You start free, with 5 summaries a month, no card required. After that, the subscription starts at 3€/month (launch offer, price locked for life as long as you stay subscribed), with 50 summaries a month included. Payment goes through Stripe, fully secure: card, Apple Pay or Google Pay. You can switch plans or cancel in a single click from your dashboard, with no need to explain yourself or write to a support desk. No fixed term, no sneaky renewal, no nasty surprise on your bill.
- Of course, and it's one of its handiest uses. Two ways to do it: either forward Abrège a voice note you received from someone else, or record a voice memo straight into your chat with it, just like you would with any contact. Either way, you get a clear summary back in seconds. Plenty of people use it to talk to themselves: dictate your ideas while walking, and get a written, structured note back. Great for prepping a report, clearing your head before a meeting, or turning a stray thought into a concrete task.
- That's up to you. By default, Abrège sends you a summary, because that's what saves time in most cases. But if you'd rather have everything, you can switch on the “Full transcription” option in your dashboard, in one click. Abrège will then send you the summary first (for speed), then the full text just below (to check a detail, a date or an exact quote). You get the best of both: the speed of a summary when a glance is enough, and the precision of a word-for-word transcription when you really need it. The setting can be changed at any time.
- On purpose, Abrège does not join groups, and that's a privacy choice. A bot sitting in a group could read everyone's messages, which we refuse on principle. So the flow stays simple: when someone drops a 3-minute voice note in the family or work group, you forward it privately to Abrège, you get the summary in seconds, and you come back to the group with the gist. Nobody else needs to install anything, and no message from the group passes through Abrège except the one you choose to send it. Handy, and respectful of everyone.
- Too long, didn't read: we save you time on WhatsApp. Stop scrolling and try the bot. Cheers. 🍿
Keep reading
Other ways people use Abrège
Your voice assistant on WhatsApp. Not Meta AI, not a business chatbot.
It doesn't chat and doesn't reply for you: it summarises and transcribes the voice notes you forward it. You save hours, without changing your habits.
DiscoverWork WhatsApp, without the wasted time.
Your clients send you voice notes? Turn them into usable text, instantly.
DiscoverWhatsApp voice-note transcription: voice into text
Turn voice into text. Ideal for keeping a written record or reading your messages discreetly.
DiscoverWho's behind Abrège
“I built Abrège because I'd had enough of 5-minute voice notes to say what fits in two sentences.”
It's a tool I use every day myself, and I build it to stay simple, honest and respectful of your privacy.
Mark Hadj Hamou
Founder of Abrège
Stop listening.
Start reading.
Made in France · GDPR · audio deleted after processing