AI video generation seems fundamentally more expensive than text, not just less optimized

About this article

There’s been a lot of discussion recently about how expensive AI video generation is compared to text, and it feels like this is more than just an optimization issue. Text models work well because they compress meaning into tokens. Video doesn’t really have an equivalent abstraction yet. Current approaches have to deal with high-dimensional data across many frames, while also keeping objects and motion consistent over time. That makes the problem fundamentally heavier. Instead of predicting t...