They do that with editing software. If it’s a still image behind them, then they would need to resize the video to a corner then layer the image behind. It all happens in the editing software. It’s done differently on each program. First layer can be the image where they drag it to stretch along the whole video then the second layer is the video. If it’s another video, then they use a camera that’s pointing at them and at the same time record their desktop (most people use OBS studio) then take both videos and layer them on top of each other in their editing software and resize the video of their face and put it in a corner. Again, it’s all done in the editing software and it’s layered/done differently on each software