Models are entering the era where physics starts to matter.
Most video AIs can create pretty frames, but basic physical rules still break easily.
The gap is about stability, weight, collision, and cause and effect.
This is where the next breakthroughs will happen.
Is video AI still failing basic physics?
I dropped a glass marble into coffee and most models broke. Slow motion, floating objects, double marbles, even jelly coffee. 💀
No perfect result today, the most stable one is Sora 2 Pro.
RT if you want more model breakdowns like this.