‘the Disney catalog is too small to create a model’
https://www.youtube.com/watch?v=lBWk3lmFhE4&list=UU9rJrMVgcXTfa8xuMnbhAEA - video
https://pivottoai.libsyn.com/20250924-lionsgate-still-cant-turn-movies-into-an-ai-video-model - podcast
time: 3 min 58 sec
You must log in or # to comment.
It is kinda fun to think about the counterfactual world where this shit had worked as expected on the first try but for obvious reasons didn’t hold up in copyright court, making the entire catalogue used for training functionally public domain.
Don’t mock me dammit, I’m allowed to dream.
A year in, there’s just one problem — AI video still doesn’t work. Two people told The Wrap: [The Wrap]
The Lionsgate catalog is too small to create a model. In fact, the Disney catalog is too small to create a model.
Google Veo 3 can’t keep a scene together, keep characters consistent, or take direction — and that’s training on twenty years of YouTube.
Oh no, who could have seen this coming (except every single artist both inside and outside the film industry)
Hey I’m not an artist or really familiar with film and I also could’ve told you this
It’s not even that hard to grasp, ChatGPT is trained on all of human textual output it can be and it can’t write a coherent story, Copilot is trained on all of open-source code in existence and it can’t write a correct line of C++, obviously a film model is not gonna be any better
Yes. Haha. Art is a skill that takes decades to master, and computers are simultaneously quite clever while deeply stupid.
