Meetings and webinars
The call quality argument, settled
A selective forwarding architecture with media servers in eleven regions, simulcast on every publisher, and recordings transcribed before the meeting invite has closed.
Why calls actually stay up
Selective forwarding
Every participant publishes once and subscribes to exactly the layers their device and link can handle. No mixing, no single bottleneck.
Simulcast everywhere
Three spatial layers and three temporal layers per publisher, so a bad connection degrades one person's view, not the room.
Region pinning
Meetings are hosted in the region closest to the majority, and can be pinned for data-residency reasons.
Live transcription
Speaker-attributed, in 31 languages, available during the call and searchable afterwards.
End-to-end encryption
Insertable streams with per-participant keys for meetings that need it, at the cost of server-side recording.
Embeddable
The same media stack as a JavaScript SDK, so you can put calls inside your own product.