I've Changed my Mind About G9... Well, Kind of...

[Ethiopia]

I'm with you on that @legolas18. I see AI being used to handle the draping of cloth and hair, and doing the final render. Rather than spend huge time and power tracing a ba-zillion light rays, the AI engine would 'know' what stuff (materials etc) are supposed to look like, where shadows would go from the designated light, then just 'do it'. I'm flabbergasted even now in the early stages how good a job it does on fabrics like satin or sheers (something that Iray flubs at) and draping of materials (unlike the joke called D-Farce). Once it figures out that folks have at least, and only, five fingers, we'll be golden.

[Legolas18]

@Inversekinematics that's a weird answer. It's like saying you won't be using a gpu because you are an artists and artists only use cpu?

Or maybe I didn't explain too well, the ai, in this case, is only used to speed up computations for the render engine, not to create things that aren't there (it's predictive, not generative).

[31971207]

"It's gonna be years for SD model and UI to offer the kind of control that a real doll house can offer."
I'll predict it won't be more than 12 months before AI will be able to make renders completely indistinguible from photography, perhaps also videos. It's 90% that way already. But I don't know if that's actually a good thing.

Single person portraits maybe. The amount of additional SD learning needed to generate images of credible interactions between multiple people out of nothing is likely exponential whereas computing power growth is geometric. It took 30+ years to go from Photoshop to where we are currently. The theories behind these models long existed just nobody bothered to put into practice due to a lack of computing power at reasonable costs. Right now the so called AI generated videos are either scenes of individuals or images ran through animation simulation to get some basic interactions. The latter is still playing with dolls.

[Author]

Posts