I think the future of that tech is rather than generating an image or a video and trying to monetise it, it'll be to come up with a model that reliably tailors content for the user. Like it'll look at their adsense profile and say "ok, this guy buys a lot of fishing gear, so generate imagery of the car being used on fishing trips." And that kind of thing.

Or maybe prerender segments for ~5000 common keywords and edit them together as needed. 

 Yeah, I've been using to to generate ideas that I can refine. This isn't  a long term solution though. I thing future generations would have a challenge "refining" prompts and knowing how to expand on an interesting idea.