Documentation
Meetings in more than one language
Ten people in a call. Four speak Hindi, two English, six Arabic. Here's exactly who hears what, who reads what, and why Jaha answers the way it does.
One rule explains the rest
Everything Jaha produces is one of two kinds, and knowing which is which answers almost every question about languages.
A meeting has one audio channel and one chat thread. Ten people, one stream. Nothing in that stream can be personalized. Whatever Jaha says out loud, everyone in the room hears the same thing, so something has to decide which language it's in.
Everything else is made for exactly one reader: the summary you open, the digest in your inbox, the notification when someone asks for you, and the dashboard itself. Those follow your own language, with no trade-off.
Jaha answers in the language you asked in
Ask in Hindi and the spoken answer comes back in Hindi, even in a room of six Arabic speakers and whatever anyone has set in their profile.
The obvious alternative is to follow the room’s majority, and it's worse than it sounds. The person who asked would get an answer they can't read, in a language chosen by five people who didn't ask. Nobody would pick that on purpose, but “use the room’s language” ends up there by accident.
When the recording can't tell which language a question was in (a short question gives it very little to go on), Jaha falls back to the language you've said you speak, then your organization’s default, then English. It never guesses without telling you: the choice and the reason are recorded on the meeting.
The room hears one language and reads the others
The spoken answer goes out right away, in the asker’s language. A moment later the same answer appears in the meeting chat in the other languages people in the room speak: one message, with each block labeled in that language’s own name.
Two things make that dependable. The translations never hold up the spoken answer. They follow shortly after, because a pause after someone asks a question is what makes an assistant feel like an obstacle. And a translation restates the answer already given instead of answering again, so the written version can't contradict the spoken one on a number or a date.
A meeting where everyone shares a language gets no translations and costs nothing extra. The extra work only happens in rooms that need it.
Setting the language you speak
In Settings, “The language I speak” describes you. It isn't an instruction to Jaha. It does three things: it helps speech recognition hear you correctly, it sets the language of anything written only for you, and it settles the choice when a spoken question is too short to identify.
Leaving it blank is a valid choice. It means your organization’s default applies and what Jaha hears takes the lead, which often works better for people who switch between languages.
Separately, the dashboard language is yours alone and changes nothing about meetings. An admin can use the product in French while their organization meets in Arabic.
One recap, translated into every language
The summary, decisions and action items are written once, from the recording, in the language the meeting was held in. Every other language is a translation of that version.
This matters more than it sounds. If each language were produced separately, two colleagues could open the same meeting in two languages and see different figures, with nothing on the page to say which is right. Instead, every version traces back to one source, and the page tells you which one it is.
Quotes are the exception that proves the rule. A highlighted quote is something a person really said, so the original words are kept exactly and the translation sits next to them. Jaha won't put translated words in anyone’s mouth.
Northwind Trading: quarterly review
Thursday 12 March, 14:00
Summary
العربية
The recap as a page in the meeting’s own language, translated for anyone else who was there.
What doesn't work
Jaha works end to end in five languages: English, Arabic, French, German and Hindi. It recognizes speech, speaks aloud, and writes answers and recaps in each of them.
Other languages aren't offered, and that's a deliberate choice. A sixth language would need a voice, speech recognition and translation we can stand behind. Listing one before all three exist would mean transcribing a meeting in the wrong language while the settings screen said otherwise.
There's also no private audio channel. Jaha can't speak one language into one person’s ear and another into someone else’s. A Teams call has a single stream, so the written versions in the meeting chat are how this works, by design.