In-person meetings,
with speaker names:
on the roadmap.
Planned, not built: a recording from the table or a phone into a bridge becomes a meeting with speaker names.
What would change in the transcript
Today the room is one voice. Planned separation and the invite list would make each person a named speaker.
Today
the room device is one participant- 10:02Board roomLet’s start with the Q3 allocation. The mandate letter came back from the client on Friday.
- 10:02Board roomTheir concern was the concentration in the two UK names. I said we would revisit the weighting.
- 10:03Priya Nairjoined from her laptopFrom my side the compliance flag is the leverage figure, not the weighting.
- 10:03Board roomAgreed, we take it to the risk committee on Thursday.
Planned, with the items on this page
separated, then named from the invite- 10:02Speaker 1 → Tom Ashbyfrom the inviteLet’s start with the Q3 allocation. The mandate letter came back from the client on Friday.
- 10:02Speaker 2 → Lena Royfrom the inviteTheir concern was the concentration in the two UK names. I said we would revisit the weighting.
- 10:03Priya Nairjoined from her laptopFrom my side the compliance flag is the leverage figure, not the weighting.
- 10:03Speaker 1 → Tom AshbyAgreed, we take it to the risk committee on Thursday.
The five items and where each one stands
Each item is planned, not built. Sizes are classes, small (S) to large (L), never dates.
1Upload a recording of an in-person meeting
A file recorded at the table becomes a meeting, through the same API as the calls.
RoadmapSpecified. No code yet.Size: size pending2Dial-in capture
A phone dials into a conference bridge, and what it hears is captured.
RoadmapCode exists on GitHub. Not merged into a release.Size: M3Speaker separation on one microphone
Voices on one mixed track are cut into numbered speakers. It does not name anyone.
RoadmapSpecified. No code yet.Size: M4Names from the invite list
Separated speakers matched to a roster or invite you supply. No voiceprints, no voice biometrics.
RoadmapSpecified. No code yet.Size: S5Room devices labelled as rooms
A conference-room device is labelled as a room, never as a person.
RoadmapCode exists on GitHub. Not merged into a release.Size: S
Need this for your board or client meetings?
See it with the founder.
Or fund it
A sponsor sets the order of the queue, co-writes the acceptance criteria and is named on the release note. Per item or per stage, agreed in writing before work begins. Not exclusivity, not a private build.
When it’s delivered, it ships open source under Apache-2.0, for everyone.
15 minutes with the founderQuestions
Can Vexa capture in-person meetings?
Not today. Vexa captures meetings on Google Meet, Microsoft Teams and Zoom by sending a bot. A meeting held around a table is not captured. Admitting a recording made in the room as a meeting is the first item on this page: specified, not started, built when a sponsor funds it.
Can Vexa tell people apart in a meeting room?
Not today, and never by recognising anyone. A conference-room device joins a call as one participant, so every voice in the room arrives on one track. Speaker separation would cut that track into distinct speakers, and name mapping would match them to a roster you supply. Both are planned. Vexa uses no voiceprints and no voice biometrics.
What does sponsoring an item mean?
You fund an item. It moves to the front of the queue, you co-write the acceptance criteria, your name goes on the release note, and it ships upstream under Apache-2.0 where everyone gets it. It is not exclusivity, not a private build, and no date is promised before the item is ordered.
Dmitry Grankin, founder, Vexa