[PolicyRecordingBot] How dominant speaker concept work in Graph communication media bot SDK?

Hello,

As per my almost a year of experience handling calling bots, let me answer this:

There is an event receiver on an audio socket to which you can subscribe in your code something like below:

private void OnDominantSpeakerChanged(object sender, DominantSpeakerChangedEventArgs e)
        {
            this.GraphLogger.Info($"[{this.Call.Id}:OnDominantSpeakerChanged(DominantSpeaker={e.CurrentDominantSpeaker})]");

            if (e.CurrentDominantSpeaker != DominantSpeakerNone)
            {
                IParticipant participant = this.GetParticipantFromMSI(e.CurrentDominantSpeaker);
                var participantDetails = participant?.Resource?.Info?.Identity?.User;
                if (participantDetails != null)
                {

                    // we want to force the video subscription on dominant speaker events
                    //// video code commented
                    // this.SubscribeToParticipantVideo(participant, forceSubscribe: true);
                }
            }
        }

You can change the above code as per your business needs.
After the bot joins the call, if let's say no one talks for 5 seconds, this event will NOT be called
Let's say you spoke for one second after that, the above event shall be called with your participant details including MSI
You will then do your business logic for said participant
After a second you stopped talking, Microsoft will again call OnDominantSpeakerChanged saying that now there is NO dominant speaker. Which means e.CurrentDominantSpeaker will be DominantSpeakerNone
Talking about media stream, you will keep getting a stream of ALL subscribed media (no matter if it's dominant or not). You have to write the logic of let's say ignoring another participant if you don't want their streams. You can also write the logic of subscribing to the ONLY dominant speakers and do not subscribe to others. Something like below
```
        ```
```
if (participantDetails != null && (participantDetails?.Id != this.CurrentUserId || this.CurrentUserId == null)) { this.CurrentUserId = participant?.Resource?.Info?.Identity?.User?.Id;
```
                // we want to force the video subscription on dominant speaker events
                this.SubscribeToParticipantVideo(participant, forceSubscribe: true);

                if (oldParticipant != default && oldParticipantMsi != e.CurrentDominantSpeaker)
                {
                    this.UnsubscribeFromParticipantVideo(oldParticipant);
                }
            }
```

Then those media streams quality can be different speaker to speaker based on camera, mic and bandwidth quality. how to manage that?

Yes, we inform to call what we want as a resolution using this line this.BotMediaStream.Subscribe(MediaType.Video, msi, VideoResolution.HD720p, socketId); however this is NOT guaranteed as it depends on the participant also. For example, if the participant has low bandwidth, the frames will be small in size. You will have to write a bridge to handle this received data and what you want.

microsoftgraph / microsoft-graph-comms-samples

[PolicyRecordingBot] How dominant speaker concept work in Graph communication media bot SDK? #329