Thanks and totally agree - especially since we're trying to enable frictionless communication, we knew we had to minimize latency from the get-go.
We use the standard WebRTC model where 1:1 Sidekicks are peer-to-peer (video streams sent directly to other clients), and 3+ Sidekicks go through SFUs that we host on AWS. These SFUs deployed all around in the world (in every major AWS region), and we assign Sidekicks to the SFU nearest to everyone on their team.
Internally the metric we aim for is to keep the end-to-end latency under 200ms.
We use the standard WebRTC model where 1:1 Sidekicks are peer-to-peer (video streams sent directly to other clients), and 3+ Sidekicks go through SFUs that we host on AWS. These SFUs deployed all around in the world (in every major AWS region), and we assign Sidekicks to the SFU nearest to everyone on their team.
Internally the metric we aim for is to keep the end-to-end latency under 200ms.