But this also opens the door to a Mickey Mouse type of situation. Perhaps in 50 years Bruce Willis will be “its” own franchise.
I suspect this will reduce earning power of stars as much as new actors.
Do you think acting talent matters, or will it be supplied by neural networks?
Suppose acting talent does matter. Then Leo and Bruce will probably be in movies 40 years from now, but they'll be puppeted by young actors, and the movies will be successful or not according to whether those young actors are good or not. Maybe we'll have a five-year period where Leo is in every movie before moviegoers get tired of it. But the young puppetmasters behind the scenes who are recording the facial expressions Leo makes in the emotional extreme closeups will still be in a good position to negotiate rich compensation deals.
On the other hand, suppose acting talent doesn't matter. Remember that, just as male porn models used to be selected for their ability to retain an erection in a high-stress situation before Viagra, Bruce's face was selected for being on a decent actor. Surely there are many more faces that are currently unknown, or painted from scratch, that would be better-looking than Bruce, that are currently not appearing in the movies because they have the misfortune of being attached to people whose talent as actors is zilch. Or to nobody at all. Moviemakers could rent those faces more cheaply.
Maybe instead of going to see movies with Leo DiCaprio like their grandparents do, teenagers will rebel by going to movies whose stars' faces are the visual equivalent of punk rock and heavy metal: a congeries of multicolored eyes melting into a vaguely vulva-shaped head, or the uncannily precisely scanned faces of supercentenarians near death, or the bare skull of a unicorn. Or, DiCaprio having been canceled for his insensitive remarks about Tiananmen Square in 02049, the teenagers will want to go see movies starring someone they love and respect, like Former President Chloe Trump.
Or maybe 40 physical years from now you and I will look back on this conversation through tens of thousands of years of subjective history (32000 for me, 44000 for you, because you've been able to afford more cloud resources) and comment on how strange it was that biological humans used to go watch movies.
I think mannerisms are going to be easy to copy as well, probably easier than the looks. It will all boil down to the technical people behind the CGI , or (worse) in the talent of the younger actors who are doing the modeling for the old actor's CGI face.
I think people underestimate how much visual arts will change when almost everyone can create whatever they can imagine. Branding is going to be the only differentiator, as in fashion or abstract art. And most fashion brands are century old
I was surprised to hear that most fashion brands were a century old, so I checked:
Tiffany & Co.: 01837
Hermès: 01837
Louis Vuitton: 01854
Chanel: 01910
Gucci: 01921
Adidas: 01924
H&M: 01947
Nike: 01971
Zara: 01975
Lululemon: 01998
Zalando: 02008
It looks like a bit less than half of major fashion brands are more than a century old, and I couldn't find any more than two centuries old.
What if people's imaginations are the limit? That seems to be the case in written fiction, for example — anybody can create anything they can imagine already, they don't have to get a CGI budget approved, or justify their location scouting expenses to an executive producer, or sign a big star to play the leading lady. And certainly there's a strong branding or fashion component to people's fiction-reading choices; famous authors tend to sell many more books. But there are plenty of bestsellers that are "first novels", and famous authors tend to imagine things most people can't, so maybe it isn't as purely branding-driven as your model predicts.
Why? There’s no guarantee a deepfake performance will be any good.