The OpenVidu team is starting to design a new feature to implement huge sessions. Basically we want to provide a very easy way to show only the participant who is speaking (dominant speaker). In that way, no matter the number of participants in the session as any participant will only have to render one video stream. Also, audio of all speakers will be combined in the server.
We plan to implement this feature in OpenVidu PRO edition.
If you are interested on it, please comment here your use case and what are the features you would use.
Your requests are more related with a controlled video conf that specifically related to dominant speaker automatic selection. Please open a new entry in the forum for such features.