🧵 Untitled Thread
Anonymous at Mon, 15 Aug 2022 06:37:02 UTC No. 913894
Why isn't this talked about more? Realistic self collision. Or as they call it "implicit skinning"
https://www.youtube.com/watch?v=GyO
There's only this one video with 64k views. All the other videos on the topic have less than a thousand views. And this was showcased back in 2014. Why is no one making something like this in Blender? Why don't people what better self collision methods? Everyone wants to talk about weight painting and softbody. As if weight painting doesn't have a ton of limitations. And as if softbody isn't a janky mess. It's like I'm taking crazy pills. Where's the implicit skinning?
Anonymous at Mon, 15 Aug 2022 07:07:47 UTC No. 913896
Because it's a performance killer.
Anonymous at Mon, 15 Aug 2022 08:54:06 UTC No. 913903
>>913896
Cloth sim is also a performance killer, but it is still used extensively. This can save weeks of work for animators and riggers, even if it would run on 1 fps.
Anonymous at Mon, 15 Aug 2022 09:45:48 UTC No. 913906
i was looking through papers that cite this and found a paper about rigging booba hehe
https://arxiv.org/pdf/2105.00763v2.
Anonymous at Mon, 15 Aug 2022 10:41:05 UTC No. 913911
>>913894
Ive always wondered why companies like Autodesk dont implement these solutions into their core packages. All game engines and base package DCC sofware still uses the decades old linear skin approach. Why are almost all other areas of 3D getting constant industry-wide updates like in lighting/rendering, texturing, and so on. But skin computation still runs on a model from the 80s or so.
Anonymous at Mon, 15 Aug 2022 10:43:50 UTC No. 913913
>>913911
because constantly iterated on commercial and in house muscle systems with FEM exist like ZIVA
Anonymous at Mon, 15 Aug 2022 13:06:30 UTC No. 913924
>>913894
all those tris
Anonymous at Mon, 15 Aug 2022 13:28:16 UTC No. 913925
>>913894
the hell is that topology
Anonymous at Mon, 15 Aug 2022 18:14:19 UTC No. 913955
>>913911
There's a lot of these algorithms in the skinning and blendshape space and what you want depends on what you're doing. Studios that care have people to implement what they need, everyone else just uses linear solutions because its good enough.
Anonymous at Mon, 15 Aug 2022 22:46:40 UTC No. 913988
>>913955
>everyone else just uses linear solutions because its good enough.
It's really not. With the way graphics are advancing, the lack of squish is more and more evident. There needs to be a sort of "standard" solution to this issue. It's crazy to me that people just accept objects intersecting into themselves.
Anonymous at Mon, 15 Aug 2022 23:15:07 UTC No. 913993
>>913988
thats just not true. Have a look at a trailer for say "Wh40k : DarkTide". There is just so much going on, that you never even notice. Same for any other game.
Anonymous at Mon, 15 Aug 2022 23:40:05 UTC No. 913996
>>913993
I'm not allow WH40K to touch my youtube watch history, but I take your point. Still, go look at any close-up of a character. Take Final Fantasy VII remake for example. The bend of their limbs are very doll like. Look at the ugly way her arm intersects her back, and how the muscle that bridges her boob and shoulder is very stiff. They do what they can to try and disguise it. But it just looks wrong, because that area should be squished together and creasing. Yet it simple is not. Please don't make this a fetish thing. The armpit area is just a clear example because of how complex it is, and how much self collision is needed to make it work.
Anonymous at Mon, 15 Aug 2022 23:47:03 UTC No. 913998
>>913996
this is a clear case where they could have used ziva RT, like they did in the spiderman game
Anonymous at Mon, 15 Aug 2022 23:54:26 UTC No. 914003
>>913998
Eh? What does Ziva do exactly? A cursory search brings up a bunch of marketing terminology about machine learning and doesn't really explain anything.
Anonymous at Tue, 16 Aug 2022 00:04:37 UTC No. 914009
>>914003
ziva is muscle sim and ziva rt is muscle sim for games in realtime
Anonymous at Tue, 16 Aug 2022 00:13:33 UTC No. 914014
>>914009
I was going to ask you to expound. But never mind. I found a better explanation. https://cgsociety.org/news/article/
Seems legit. Although, Ziva RT seems to only apply to faces at the moment. I could be wrong. It's bothersome to sift through all the marketing jargon.
Anonymous at Tue, 16 Aug 2022 00:15:14 UTC No. 914015
>>914014
>>914009
Oh, but learning about Ziva RT doesn't quench my desire for a more standardized self colliding skinning solution. Something more intuitive and lighter than cloth, capable of working in real time.
Anonymous at Tue, 16 Aug 2022 00:20:02 UTC No. 914017
>>914014
Ziva rt is separate from ziva face. Ziva rt was used for spidermans suit in the game
Anonymous at Tue, 16 Aug 2022 01:09:42 UTC No. 914032
>>913911
Because not everything needs to be skinned with a computationally heavy method like muscle sims you retarded zivashill, kys
Anonymous at Tue, 16 Aug 2022 10:08:09 UTC No. 914072
>>914032
im stating the facts that in one of the biggest games in the world they "chose" to use ziva rt
Anonymous at Tue, 16 Aug 2022 13:14:16 UTC No. 914098
>>914032
wow looks like I stepped on someones tail. whats with the hate brah?
Anonymous at Tue, 16 Aug 2022 13:21:14 UTC No. 914100
https://www.youtube.com/watch?v=ViI
>>913894
>Why isn't this talked about more? Realistic self collision.
In the 3D industry there are two types of people: artists and coders. Stuff like "implicit skinning" are created by coders with no real understanding of pipelines, artistry, resource management, etc.
As a general rule: if the presentation is not within a viewport of a 3D software like Maya/Blender/3DSMAX just ignore it. Academic papers are written for the sake of writing them. They serve no purpose whatsoever for people like us.
Anonymous at Wed, 17 Aug 2022 01:28:53 UTC No. 914270
>>914100
That video is very impressive. And I've always figured something like that is possible using only bones. But I'm still relatively a beginner, and so such a bone setup would take me many months to figure out. Shame the guy didn't show his bones. If he did, then we could at least backwards engineer how he did it.
>As a general rule: if the presentation is not within a viewport of a 3D software like Maya/Blender/3DSMAX just ignore it.
Sorry, but no. I can't ignore it. Because someone has to be the coder who makes such technique viable for artists. And if artists to make a big fuss about what we want, then coders will continue not bothering to develop what we want. They feed us excuses like "it's a computational nightmare", and we just accept that? We accept that something CRUCIAL to our artistic endeavors is simply too hard to do? Bull fucking shit. Someone has to figure out a way. But they never will, if they don't put their minds to it. If they just brush it off as a non-issue.
Anonymous at Wed, 17 Aug 2022 02:07:00 UTC No. 914279
>>914270
>Because someone has to be the coder who makes such technique viable for artists.
That's what I meant. Coders have no vision and it's the artists who ask them to program stuff that they need. This is why when a coder does stuff on their own nobody uses it. They understand the "how" of making things but they don't have the vision to see the "why".
I started as a 2D artist, got into 3D and now I'm also coding. What I learned is that coders code and that's it. There are software developers out there who have no idea how to code but they have the vision. Coders have no initiative or ambition, they just want to code, get money, and play videogames. That's it. The guy who created GANs did it as a dissertation paper and gave it away for free. He didn't have the foresight to realize that his little invention will change the CGI landscape completely.
Anonymous at Wed, 17 Aug 2022 03:58:39 UTC No. 914301
This somewhat similar to a muscle system, which are used extensively in cg animals.
UE5 has something like this with machine learning powered blendshapes and it works in realtime.