With all of the AR capabilities of actual phones, I wonder if there is any app to record video with all the spatial data of the camera which lets you export the virtual 3D camera directly from the shot to be used for compositing.
This is the only example I've found. It uses the Unreal game engine to generate the realtime 3D and you need a hefty graphics card to make it work.
I asked a similar question here: Budget Mandalorian backgrounds
The only one I know of is CV-AR from cineversity, which uses apple AR kit to capture facial animation, and works in conjunction with Maxon Cinema 4D. I don't think depth sensing on phones is quite to the point where anything further than a selfie is very accurate, although I'm sure that'll change in the very near term if it hasn't already.