Meeting Intelligence
LiveRecord or upload a meeting, get a transcript and minutes, turn the action items into tasks — with the speakers labelled, never identified.
- ✓Record in the app or upload a recording, up to 300 MB (about five hours)
- ✓Live transcription in ~45-second chunks, polled onto the screen while the meeting runs
- ✓Pause and resume, with the pause and its length drawn into the transcript and noted on the export
- ✓Speaker labels — Speaker 1, Speaker 2 — inferred by a model and renameable; never voice identification
What it is.
Record a meeting in the app or upload a recording afterwards — up to 300 MB, about five hours of audio. A live recording goes up in chunks of roughly 45 seconds and each is transcribed as it arrives, so the transcript is on screen while the meeting is still running. Pause and resume are part of the record: the transcript shows where the pause fell and how long it lasted. When you stop, the minutes are generated once from the whole transcript — a summary, the decisions, the discussion, action items, unresolved questions, commitments and next steps.
Speakers are labelled Speaker 1, Speaker 2 and so on. Those labels are inferred from the words by a model, never identified from a voice, and every screen says so; you can put a person's name on a label, and it stays marked as inferred. Action items become tasks in the same task list the rest of the product uses, on your click — nothing is filed silently. Minutes export to PDF or Word, and a meeting linked to a matter writes its record onto that matter's timeline: meeting started, minutes generated, three tasks created.
Three steps.
End to end.
Start a recording on the phone or the desk, or upload a file after the fact. A meeting left paused for four hours, or recording with no audio for six, is closed by the system and finalised with what it has.
Chunks are transcribed as they arrive and the screen polls for the next one. On stop, the transcript is labelled by speaker and one model call produces the minutes. A meeting that fails says so, with a reason and a retry — it does not spin.
Name the inferred speakers, correct the minutes, create tasks from the action items, export to PDF or Word, or ask Vio to draft the follow-up email — which it shows you and never sends.
What you get.
- ✓Record in the app or upload a recording, up to 300 MB (about five hours)
- ✓Live transcription in ~45-second chunks, polled onto the screen while the meeting runs
- ✓Pause and resume, with the pause and its length drawn into the transcript and noted on the export
- ✓Speaker labels — Speaker 1, Speaker 2 — inferred by a model and renameable; never voice identification
- ✓Minutes generated once from the transcript: summary, decisions, discussion, action items, unresolved questions, commitments, next steps
- ✓Minutes can be corrected by hand; the record then says who edited them and when
- ✓Action items filed as tasks in the shared task list, once each, on your click
- ✓Export to PDF or Word, dated in your own timezone, with every speaker label marked inferred
- ✓Linked to a matter, so the meeting, its minutes, its tasks and its exports appear on the matter timeline
- ✓Search a phrase across every transcript and every set of minutes in the organisation
- ✓Abandoned recordings closed by the system — four hours paused or six hours silent — and finalised with what was recorded
- ✓Transcription runs on OpenAI's speech model, disclosed in the privacy policy; unavailable in an air-gapped deployment, where the Record button is disabled
Quick answers.
No. The transcript is split into turns and a model labels them Speaker 1, Speaker 2 and so on from the words alone — there is no voice identification. You can put a name on a label, and the export still marks that label as inferred, because the name is your assertion rather than the system's.
Yes. Each section can be corrected after generation, and the meeting then records who edited it and when — the provenance line reads AI-generated, edited, rather than pretending the model wrote the final text. You can also regenerate the minutes from the transcript, and an action item that already became a task keeps its task through a regeneration.
No. It records from the device's microphone or takes an uploaded file. There is no calendar integration and no bot that joins a call.
No. Transcription is priced per minute of audio, not per token, so it is not in the token log. The record of it is the meeting itself — the recorded duration and the transcription model used.