Looking to find 4 or 6 fingers 😉

Okay, that's the best AI-generated video I’ve seen in a while. Huge improvements in movement and photorealism. But I’m guessing it’s currently limited to around 3–5 seconds per sequence, since it still struggles with consistently portraying the same "person" across angles or with fast movements.

From what I understand, that’s because most current models generate each frame individually rather than truly understanding motion or character consistency across time. So things like identity shifts or unnatural transitions are still a challenge.

That said, if this is the best we’ve got right now—imagine where we’ll be in 3 years.

Either it’s going to be amazing, where anyone can create full videos or even “movies” on demand… or it’s going to be a total nightmare.