Basically, the camcorder object has a 3D array, and a trigger collider in it. 24 times a second, it gets the Collider2Ds of everything it’s currently colliding with, converts them into meshes, and get its vertices. After that, it works out which vertices are within the bounds of the camcorders’ collider and uses that data to work out the vertices of a new mesh that completely overlaps with the camcorder. It then stores the data in the vector3 array as [Time the things were in there][Collider][Colliders vertices].
Then, in a separate script, it has an int corresponding with the time to create, and it creates an arbitrary mesh corresponding with the different objects, and their vertices. It then creates a collider for each of the objects by setting the points of a polygon collider 2D to match the vertices of the object.
I really hope this was coherent, lol.