This is not for games. Realtime is different than "for games". Realtime means the cloth simulation is realtime, but this doesn't mean that your engine can actually handle more than what you see in this video (sim + enviro + light + interactions + models + anims etc...)
To simply reframe it from a current gen perspective, you know that we can do cloth sim now but :
- Do you see every "main" character having big sims on them in general? No
- Do you see a lot of NPCs with cloth sims? No
- Do you see a lot of cloth sims on enviros? No
Now, combine that state of things with : we have to rely on DLSS and Framegen to just display the games at 4K, and most games nowadays release in abysmal states with poor to very poor performances. Highly unlikely you'd prioritize cloth sim.
But let's imagine the cloth sim algos were like, dividing their overhead by 2 tomorrow, for game engines this time. This does not mean we even have room to actually make a big use of it. As devs, we have already a massive amount of work optimization wise without accounting for clothes. It's highly unlikely we'll allocate 1 or 2 ms just for a couple of jeans and skirts moving around. So most sims will be "low details" at best.
When you see that VAM1 still fucks up any CPU even if we know why. Pushing that sort of high end sim without accounting for all the other physics computation, you can easily picture a game running at 3fps.
These are research papers/techs, the delay between this and real life application and then game application can be long.
Also, game engines don't really prioritize physics in their roadmap. We had to wait years for Chaos in Unreal Engine, and it's total mess.
