Translated captions, live translation and an AI copilot during the meeting: how each participant gets them inside an InterMIND room (2026)

A call where two or three languages are spoken has one question that cannot wait for the notes: what is being said right now, in words I read without effort? Two answers are on the market. One is a copilot that runs beside the call — on the listener's own screen — and shows that person a live transcript, translated, with AI help while the conversation runs. The other is a room where the host provides the same thing for everyone in it, and where it stays after the call.

This guide shows the second answer in detail, as it works in InterMIND today, after a short map of the first.

The copilot beside the call: what Bearbits states

Bearbits was shown on Hacker News on 4 October 2026 by its author, who built it "mainly for myself and my wife" to "see in real time what's being said, irrespective of the language being spoken", with real-time translation into a chosen language and "suggestions for what to say that are grounded in actual web search". The public pages state the following (checked October 2026; sources at the end):

• Where it runs. "Bearbits runs in your browser and captures audio you share" — it "never joins your call as a participant" and works with "Zoom, Google Meet, Teams, or any browser-based call". The documentation describes capturing "microphone and system audio", with a tab or screen share used only for the audio; no recording is kept.
• Languages. "Live multilingual transcription with speaker labels" in "60+ languages", and "Real-time translation into a selected language". On language changes: "Set language hints so the recognizer matches the languages you expect. You can adjust during a meeting if speakers switch languages." Transcription runs on Soniox, web search on Exa.ai (per the author's post).
• The copilot. An in-call assistant with "What do I say?, Anti-BS, Catchup, and Ask AI — grounded in your transcript and workspace context", and "Current web research with public sources during the meeting".
• Who sees it. Transcripts are "private to your authenticated account". Whether other participants of the call see anything is not stated in the public documentation.
• Price. Prepaid credits, "No subscription": $15 for 5, $45 for 20, $100 for 100; "Each credit covers up to two hours, and stays ready for 90 days." The author offers "30 minutes usage for free".

That is the copilot: one participant's own window beside somebody else's meeting, paid per meeting by that participant.

The room: the host provides it for everyone

In InterMIND the meeting itself is the place where every participant speaks, hears and reads in their own language. Nobody brings a tool to the call — the host opens the room, and what follows is there for every person who joins the link, with or without an account.

Captions and voice, each in the listener's language

Translation in a room is switched on once the room has at least two languages — More → Enable translation, or Alt+T. From then on each participant hears the others translated into their own language, in a voice modelled on the original speaker's rather than a generic narrator (how that pipeline works). A five-person call with five languages is five simultaneous translations, each picked by its listener.

The captions are the text of that same translated stream. More → Show subtitles puts a banner at the bottom of the screen with what the current speaker is saying, in the language you chose; the banner fades when nobody speaks, flips to right-to-left when the target language calls for it, and remembers its on/off state per browser, so the choice made in one meeting holds in the next. Two limits, stated plainly:

• The subtitle banner shows the translated text of the current speaker. It does not print a speaker label next to each line — speaker attribution lives in the transcript after the call (below).
• The banner is text on top of the translated audio, not a separate transcript window; the live written record of the meeting as a whole is produced when the meeting ends.

Languages, and what happens when someone switches

Voice, chat and shared notes share one list of 23 languages; files shared into the chat use a wider list of 30 (why the numbers differ). Every language is available on every plan.

Each participant has one language of their own: by default it follows the interface language, and the Translation language submenu in the More menu changes it at any moment during the meeting — the switch takes effect immediately. That is also the honest answer to the mid-meeting switch: InterMIND does not detect on its own that a speaker has moved from Spanish to English in the middle of a call. The speaker's language is the one they declared, and changing it is one click in the same menu — speech recognition is given that declared language rather than guessing one.

What a guest needs to install: nothing

A participant opens the link in a browser, picks their language on the join page, and is in the room with translation, captions and their own words going into the shared transcript — no account, no payment surface. Apps for the desktop, iOS and Android exist for people who use InterMIND every day; the guest of one meeting needs none of them. The host's plan pays for the room; participants pay nothing, whoever hosts.

The transcript and the recap: the part that stays

Transcription has no Start button. From the first word of a live meeting, speech is captured on the server, attributed to the person who said it, in the language it was spoken. When the meeting ends, a recap appears in the meeting chat — the summary in each member's own language — and its Transcript tab holds the full speaker-attributed, timestamped transcript, translated on demand into the language of whoever opens it. Every signed-in member receives it; a channel keeps every session's recap next to its chat.

The limit here is the one the terminology hides: a guest who joined without an account gets the full room during the call and nothing afterwards — there is no account to deliver a recap to, and the anonymous identity is not kept once the call ends. An account is free, and it is what makes the after-the-call half reachable.

The AI in the room

Mia is the organization's AI, and she is in the same window as the meeting. Signed in on a paid plan, Cmd/Ctrl + / (or Mia in the sidebar) opens her panel on any page and inside a meeting room. She knows which screen you are on, answers from your organization's meetings, chats and files, and can prepare a meeting link or a message for you to confirm. When a question needs facts from outside the organization — a company, a price, a law, another product's documentation — she searches the web and gives the address of every fact she takes from it. What leaves InterMIND for that search is the phrase Mia words herself, never a document, recap or chat; an owner or admin switches the search off for the whole organization under Settings → AI features. Aloud in a meeting she does not search.

Every AI feature of the organization — recaps, document summaries, the writing assistant, Mia — runs on one language-model gateway the organization chooses: Azure OpenAI in the EU Data Zone by default, Google Vertex AI in the EU, Amazon Bedrock in the organization's own AWS account, or the organization's own OpenAI-compatible endpoint; AI can be switched off per organization, and meetings are still transcribed and translated with it off (the gateway, in detail).

See it instead of reading about it

• Try the live demo — no signup: speak, and hear the room translated back to you with the captions on.
• See the benchmark — per-language-pair translation quality, measured monthly on real traffic, methodology public.
• Read the subtitles docs and the transcription docs — the full mechanics.

FAQ

How do I turn on translated captions in a meeting? In an InterMIND room, open More in the control bar: Enable translation (or Alt+T) once the room has at least two languages, then Show subtitles. The banner shows what the current speaker is saying, translated into the language you chose; the same item becomes Hide subtitles. The on/off state is remembered per browser for the next meeting.

Can the captions be translated into a different language for each participant? Yes. Each participant picks their own translation language — the one setting covers the translated voice, the captions and chat messages — so in one room the same sentence appears in Spanish on one screen and in German on another. The choice is per person, not per meeting, and changes at any moment from the Translation language submenu.

Is live translation or captioning an add-on I have to buy? No. Translation, subtitles and automatic transcription are part of every InterMIND plan, the free one included; what the plan sets is the room size and the monthly translation time, listed on the Usage & Limits page and on Pricing. Participants never pay — only the host's plan counts.

What is the difference between live captions and a transcript? Captions are the translated text of the current speaker, shown on screen while the meeting runs, without a speaker label. The transcript is the written record produced when the meeting ends: every turn attributed to its speaker, timestamped, kept in the language it was spoken and translated on demand when a member opens it in their own language. It is reached from the recap in the meeting chat.

Does the AI help during the call, or only after it? Both, in different forms. During the meeting a signed-in member on a paid plan opens Mia with Cmd/Ctrl + / in the same window: she answers from the organization's meetings, chats and files and searches the web when a question needs it, with a source for every fact. After the meeting the recap and the transcript arrive in each member's language. A spoken Mia who listens in the room and answers by name is in preview — it is being tested inside InterMIND and is not switched on for customer organizations yet.

Does a guest need to install anything to get the captions? No. A guest opens the meeting link in a browser, picks a language on the join page, and has the translated voice, the captions and their own words in the shared transcript. Apps exist for desktop and phones, but a guest of one meeting does not need them — and pays nothing.

What happens if a speaker switches languages in the middle of the meeting? The speaker changes their declared language in More → Translation language; the switch takes effect immediately. InterMIND does not guess a language change on its own: speech is recognized in the language the speaker declared.

Sources: Bearbits — home page, Bearbits — documentation, Bearbits — Live meetings, Bearbits — pricing, Show HN: Bearbits – AI copilot that helps during meetings instead of only after, all checked October 2026. InterMIND facts: real-time translation docs, live subtitles docs, choosing languages, transcription docs, meeting recap, AI features and Mia, usage and limits, language breakdown, own-voice translation, live benchmark.