Security and privacy
Content of a meeting is confidential. The design says so.
Our visitors are executives and investors, and many sit behind national filtering. That shapes the architecture more than any feature does.
Translation never sits in the path of your voice
Audio and video travel on one route; translation listens on another. If a translation service slows down or stops, the meeting carries on and only the subtitles pause.
Audio is not kept
The voice is held in memory, sent to be transcribed, and discarded. Nothing is written to disk unless the speaker has explicitly turned recording on.
Keys stay on the server
Speech recognition, translation and speech synthesis are called from our servers only. The app never talks to those providers, and no provider key is ever shipped in it.
Notifications carry no content
“Your meeting has started” can be a notification. What was said, or its translation, never is — it does not leave our systems in a push message.
One origin, over TLS
The API, realtime channel, media and audio clips all come from our own domain on port 443, with our own relay for restrictive networks. No third-party CDN, no analytics, no trackers on this site.
Access is checked on the server
Creating a meeting is a permission, separate from joining one, and it is checked server-side whatever the app shows. A suspended account is cut off on its next request.
Files and secret chats
Every file is checked before it opens. Secret chats are end-to-end encrypted, and translation is off in an encrypted call — the one place we cannot read is the one place we do not try.
The model may not add anything
Translations are instructed to preserve numbers exactly as spoken and never to add units, currency or anything not in the original. A test for exactly that runs on every change.
Questions
Before you ask
Do guests need an account?
No. A guest gives a name and a meeting code and can join. Creating a meeting needs an approved account.
Which language do I speak in?
Yours. You are asked only which language you want to read or hear. The language being spoken is recognised from the voice, not filled in.
How many languages can one meeting hold?
Up to fifty people and five languages at once. Translation is done once per language, not once per person, so a crowded room is no slower than a small one.
Is my voice recorded?
No, unless the speaker turns recording on. Otherwise audio is held in memory only long enough to be transcribed.
What if translation stops working?
The meeting does not. Voice and video run on a separate path, so only the subtitles pause until translation is back.
Is there an iPhone version?
It is coming soon. Until then the browser version works on iPhone.
Hold the meeting in your own language.
The other side holds theirs. Nobody waits for a turn, and nobody is the one person in the room working in a second language.