Still think it’s a solution in search of a problem
None of these are real.
A new and very successful strategy for realistic video is combining the FLUX.1 image model with a good video model (I've been using Runway Gen-3). And then to top that off you can add Live Portrait into the mix and act out the facial expressions or dialogue to graft onto the video... the tools just keep flying in and getting better every single time.
https://v.nostr.build/BBnoRaa4mIrhVq2y.mp4
https://v.nostr.build/1FOaDHeOByQRQ8M8.mp4
https://v.nostr.build/LiquPKh8s8g5roPT.mp4
Discussion
I disagree. In many contexts its totally used as a gimmick, but as an editor and filmmaker there is nothing at all questioning in my mind for the value of these. If you've ever done any complicated VFX work you'd know how valuable this is. As a compositing tool this stuff is like fucking magic.
Much of LLMs shown and how they are proposed aren't working and I think the development direction of them is going to change. Looking at how people *actually* use them vs how everyone said they were going to be used makes me think things will start to shift soon. You aren't entirely wrong with how things are today.