I don't care how the background is generated, if the making algorithm can't understand which part is my head, which is the headphones and which is the headrest on my chair. Unless I use green screen, the auto detect for the background is too inaccurate do use any background filters.
How about AI-ing that instead?