Step-by-step nvidia GPU passthrough for cuda/vulkan compute applications

Depends on your setup. Assuming you have more than one graphics card, connect the monitor to an interface of a card that is not hidden from dom0.

If you’re using laptop your display interfaces are most likely wired to the gpu you’re hiding from dom0.

I don’t know what google meet does, but cuda is generally for pure compute, not for graphics. If you want your gpu to render things, use prime or set up a desktop environment in your gpu qube (see messages above)