I use it for the chroma keying in VC software that doesn't support it, like Meet (which has a machine learning system that tries to identify the background). I often present other windows and using OBS is often more ergonomic than connecting and presenting a second window; I can composite a transparent version of my head over what i'm presenting. Or present my whole screen, etc. Basically, twitch streaming for video games and other stuff changed how I do VC.

How do you do present window sharing to a Meet call?

Using OBS gives you a very flexible way to stack a window capture with video sources, composite them, etc, yes, but your output is then routed through the virtual camera driver and sent to Meet as a regular webcam feed.

So because Meet thinks it's just a webcam feed, your coworkers will just see your shared window as a tiny video tile downsampled on the sending end down to 720p, and the only way they have of seeing it large is to pin your video tile manually.

Is there a trick here where you got OBS to output the composite feed to a chromeless window then you share that as a regular application window presentation?