I'd guess maybe four or five years. I feel like initial proof of concepts might be out in two though. We seem to have animation (mostly on the way, but animating 3d models is making progress), voice, object recognition, camera work etc but creating 3d models is soon and after that the main thing is designing "sets" and "backgrounds" at least kinda accurately. I think architecture programs might actually help in this area, like if interior designers or architects start applying AI to their work maybe we can make breakthroughs.