Video segmentation" refers to techniques that make it possible to separate objects in the foreground of a scene from the background (in order to then treat both layers as two or more different visual layers).
Up to now, either a monochrome studio background (which is then called chroma keying) was used for this. Or you can have the corresponding digital artists manually crop the desired objects one frame at a time.
It has been possible for some time now to detect objects in a video using the AI algorithms that are currently burgeoning. But Google has now completed a video segmentation app that can even work on smartphones in real time. In contrast to a green screen image, a homogeneous background is no longer required for the app.
Objects in the foreground are reliably detected and segmented even in front of turbulent background motifs. However, the results of the edges are not yet on the level of a professional key. However, this is probably not particularly relevant for the selected application case as a fun app.
It should be noted that automated keying without a special background is now slowly but surely coming to us.
