>Nvidia's PartPacker, a new AI-driven method for generating 3D objects from a single image, allowing each part of the object to be separately edited, which is a significant advancement in 3D modeling technology.https://huggingface.co/spaces/nvidia/PartPacker
>>1013847Thread's there>>1009076
>>1013847>asymmetrical>shit tier surface qualityuseless, as is every other "AI" modeleronly NGMIs are afraid of this trash
>>1013847>https://huggingface.co/spaces/nvidia/PartPackerGonna use that rn, useful enough to get a half decent starting point for sculpting. It misses details but gets proportions just right which I am personally very bad at.
It's dogshit. Better off just using a basemesh instead of using artifact generator 2000This is like 3d scanning but awful in all ways. The bullshots on the website aren't impressive generations either.
>>1013886This guy gets it. None of you have ever used zbrush and it shows.
>>1013847Seems helpful for getting orthographic images to use as a reference
This one might be better, doesn't do segmentation but surface quality is arguably better.https://www.youtube.com/watch?v=XRFlnXeOdwwhttps://huggingface.co/spaces/ilcve21/Sparc3D
>>1014130If you have to rely on an ai to make you a basemesh that could be done in 10 minutes you should delete yourself and never open any 3d software ever again, because chances are you are a complete failure in all other aspects if you can't grasp the fundamentals
>>1014219>10 minutes>still doesn't get itLmfao cope harder any time saved is a win and you'll deny that while using automated tools within your package.
>>1014221Keep gimping yourself then
>Cris is gonna be making better models than all of us by the end of next year
>>1014214this is pretty good. doesn't do mesh segmentation though.
Didn't read didn't watch it's but it's still definitely shit
>>1013847
>>1013847What's the topology like?
>>1013847Can it make a game ready, rigged model?Because, if not, the 3d modelers and animators are still in business.
the more i learn about 3d the less impressed I get with these. What's the topology look like? All I've seen come from diffusion models are basically things you an build off of but never actually good enough for use without artist input and refinement or heavy editing to the point an experienced artist is still irreplaceable.
>>1013847>It really is over..Your mom is over at my place sucking cock, just like AI sucks donkey balls.
>>1014289Looking at this more makes me less impressed by it. Like sure it got some of the rings right and it understood that there are some barrel cylinder thingies on the back, but its nowhere near being the same model and you can see where it just gave up on lot of the details. I urge some of the anons here that are worried about it developing really fast to look at what the models were capable of in late 2023 and compare it to what we have now, and I really mean late 2023, not the first image to 3D research papers. For example this model I took from some discord server back in novermber of 2023. Yes, it is coming but not as fast as some anons here predict, and it is basically guaranteed at this point that we reach AGI within the next 20 years or humanity goes extinct.
>>1013847it ain't there yet
>>1013847the future is being able to retopo then?
>>1013847over for whom? I like making thingsit wasn't over when I could go on fiverr and pay some southeast asian to make me a model for peanuts
>>1014221This. AI is the strangest thing to take a moral stand against. All I see is more automation to make your job easier. It's not going to steal your job. It still needs a living person to give it direction.
>>1013847it didn't work very well for me.
