AI Video Still Cannot Do Hands and Nobody Admits It
Runway and Luma will happily generate a person talking. They will quietly generate six fingers if you blink.
I am not the target user. I just wanted it to work.
Runway and Luma will happily generate a person talking. They will quietly generate six fingers if you blink.
The bit that confused me: The fix is not a better prompt, it is a shorter clip and a tighter frame so the model never has to draw the hard part.
What I would do differently: For anything with a human on screen, I generate the environment and composite the person. The model is a set designer, not a casting director.
It is fine. It is not life-changing, and that is okay.
