ahalpert
Major Contributor
content aware fill, for example. I've used it successfully for extending the canvas as you've done on the portrait above.
I took a look at adobe's content-aware fill for video when it first came out and it didn't seem ready for prime time. has it improved?
and can it do objects/people? or more just like background stuff? the tricky part of the basquiat portrait above isn't the canvas but adding the rest of his arms.
I get the impression that all these tools will converge, that different AIs doing similar things perform similarly. the techniques and technology percolate into all of them. training AIs is getting cheaper, and every company will be able to make their own. the full array of functions will basically be folded into every NLE eventually

