Otter has added 13 languages to the four it already handled: English, French, Spanish and Japanese. A feature that follows a conversation as speakers move between languages is promised for later. For now, a user still picks one language in the settings for each meeting.
A one-language setting is a rule about who gets to talk
A transcription tool that needs one language chosen in advance puts a rule on the room. Anyone who speaks something else gets garbled text, or none. Teams respond in the sensible way: they pick the language everyone can get through, and the people who speak it as a second language say less, say it more simply, or stay silent.
Following a switch removes the reason for that rule. Otter is not alone in heading this way. Fireflies, another notetaker, has a multi-language mode that detects several languages in one meeting, though it is still labelled beta. When two products move in the same direction, the one-language meeting is on its way to being a setting nobody has to think about.
The person whose words matter most is often the one who switches
Take the head of a school department who meets parents. The teacher speaks English, and a parent answers in Cantonese, one of the languages available now, then goes back to English for a word like "assessment". The parent's own words are the most important part of the meeting, and a one-language notetaker is most likely to lose exactly those: a worry about a child's sleep, a medicine, a change in who collects the child.
If the notes catch both languages, the follow-up message reflects what the parent said as well as what the school said. The teacher still decides what to act on. The notes just stop dropping half of the conversation.
Test it on a real call before relying on it
Switching is not available yet, so the first test is whether it ships and works for your languages. Otter built the new language models itself, trained them on public datasets, and had volunteers who speak each language check the transcripts. The accuracy figure in its announcement covers English only.
So before a team relies on the notes, it should record one real call in its own language pair and read the places where a mistake costs money: names, quantities and dates. Those are also the words that most often switch language in the middle of a sentence.
Language coverage also decides who benefits from the live suggestions that AI coaching in meetings offers. They help the newest person most only if the tool hears the language that person actually speaks.
Within a couple of years, the language of a meeting will stop being a setting in notetakers. What changes is who talks: the people with the best information stop translating themselves before they open their mouths.