An example that shows how to connect RealSense depth camera to Unity VFX Graph
The PointCloudBaker component converts a point cloud stream sent from a RealSense device into two dynamically animated attribute maps: position map and color map. These maps can be used in the "Set Position/Color from Map" blocks in a VFX graph, in the same way as attribute maps imported from a point cache file.
Is it possible with Azure Kinect?
There is a project for you. Check out Akvfx.
Which RealSense model fits best?
I personally recommend D415 because it gives the best sample density in the current models. Please see this thread for further information.