Meeting Copilot joins Google Meet, listens through a selected transcription language without keeping an AI session open, and responds when you say “GPT” and ask. GPT-6 Astra reasons over recent discussion; GPT-Live speaks the answer.
Hi Product Hunt — we built Meeting Copilot because the most disruptive part of asking AI during a meeting is leaving the meeting.
It joins the Google Meet your team already uses and stays quiet. When someone says “GPT” and asks a question, GPT-6 Astra examines the latest discussion, surfaces the assumption or risk that matters, and hands a concise brief to GPT-Live. GPT-Live is instructed to speak the answer back inside the call in the participant's language.
We deliberately separated listening from AI connection: while the team talks, Recall streams finalized captions in the selected meeting language, but the bot does not keep an OpenAI voice session open. Each requested answer is capped at 45 seconds. We do not retain raw meeting audio or video, and stored captions can be deleted.
For this launch, we added 83 selectable transcription-language configurations, a universal name-first question format, Japanese and English product and in-meeting status UI, and the Astra reasoning layer. Japanese and English are validated end to end; the wider language surface depends on each transcription environment and is not presented as 83 fully validated voice localizations. We would especially value feedback on where a short outside perspective improves a live decision — and where it interrupts instead.
Hi Product Hunt — we built Meeting Copilot because the most disruptive part of asking AI during a meeting is leaving the meeting.
It joins the Google Meet your team already uses and stays quiet. When someone says “GPT” and asks a question, GPT-6 Astra examines the latest discussion, surfaces the assumption or risk that matters, and hands a concise brief to GPT-Live. GPT-Live is instructed to speak the answer back inside the call in the participant's language.
We deliberately separated listening from AI connection: while the team talks, Recall streams finalized captions in the selected meeting language, but the bot does not keep an OpenAI voice session open. Each requested answer is capped at 45 seconds. We do not retain raw meeting audio or video, and stored captions can be deleted.
For this launch, we added 83 selectable transcription-language configurations, a universal name-first question format, Japanese and English product and in-meeting status UI, and the Astra reasoning layer. Japanese and English are validated end to end; the wider language surface depends on each transcription environment and is not presented as 83 fully validated voice localizations. We would especially value feedback on where a short outside perspective improves a live decision — and where it interrupts instead.