Creative Chairman at Pereira O'Dell, Chief Creative Officer at Serviceplan Americas, Founder at Silverside AI, Author The Girl from Wudang, board member at The AdCouncil and The One Club for Creativity
Luma’s cinematic motion (both camera and subject moves) are a big step for AI video, breaking that “floating pause” feel every video seemed to have until now. Behind that specific improvement, there’s more evidence that we shouldn’t take for granted the limitations of the technology. Whatever they are, they may just be a matter of having the project list finally hand on it to get it solved. It happened to the multiple fingers, it reached the rovering… which makes sense if you think there are real people, but not enough, and real compute power but not enough, working on these problems. So the resources are limited and they have to go for the bigger, more frequent, more impressive or fundamental ones first. But once any problem lands on the hit list… it’s done. So don’t look at the current problems of AI as essential limitations. Most of them will be solved as soon as the project manager says it’s time. Most of them, not all, I said. The question I keep asking myself is: what are the problems that will never be solved because they are inherently part of what being AI is?
We are LOVING the generations from Luma Lab's Dream Machine. Now it’s time to answer the question, how does it compare to these other big-name video generators. For most of these shots, Luma wins. #lumadreammachine #aifilmmaking #aivideo #filmmaking
Runway hallucinates way too much for my taste. It's really difficult to use for any practical applications.
Stunning. To answer your question I think the output corresponds directly to the quality of the input. And that’s where humans will always come in.
Another example of a 10x improvement that sudfenly disrupts in a big way. We better get used to that 🚀
corporative culture has being teaching us to be all the same, fit the mold. Now, what will make professionals survive in the corporate world is exactly the opposite: their individuality. Our authenticity in a intrinsic way is what machines will never do.
Luma took a big step forward. Wow.
Aprendiz na Mirum e i-Cherry
1moI wouldn’t bet the limitations of gen video is close to be solved. This is a completely different beast compared to image and text. The way gen video is created is far different than other mediums and fixing it is not a matter of bigger models or computing power. The expectations around video generation are in complete disarray of what is needed to get to a solution that figures it in ways people are looking for.