It integrates with jitsi, which is a fairly good tried and proven solution. Meanwhile, The matrix developers are working on their own implementation of voice and video that plays a bit nicer with their room permission system. For one to one conversations, there is a turn-based solution for voice and video.
First, it’s federated, meaning that different instances of discord can talk to each other, much like Lemmy.
Second, it allows for encryption. Matrix uses the same double ratchet algorithm present in signal.
Third, joining groups is optional. This is perhaps the biggest user interface difference between discord and matrix. Each conversation exists in a independent channel, or room as they are called. Rooms can be grouped together the way you would see in discord, but they usually exist independently of the groupings. Incidentally, matrix groups are called spaces. There are edge cases where rooms are not independent from spaces, but by and large it is not something most users will have to worry about.
I actually find it nice compared to discord. It’s simple. If you need more complex moderation in matrix, there are bots you can use
Yes. Matrix uses an integrated jitsi widget for voice and video. It is unfortunately not quite as polished as discord for voice and video, but it does work.