Loading...
> >

You know what Ai would be good for in 3D art?

jdavison67jdavison67 Posts: 645
in The Commons

Phone cords! Ropes! Chains! Vines! Tubing!

Imagine if you could designate a starting point, a rough path, and an end point, and let Ai fill in the rest.

Think of it like Tweening in animation. you set key frames, and let the in-between frames get posed for you.

 

JD

Comments

  • Peter WadePeter Wade Posts: 1,622

    You don't really need AI for that, just basic physics. And a good physics engine would probably create more realistic results than an AI that would basically be making guesses based on pictures, some of which are similar to what you are trying to create.

  • TomhipTomhip Posts: 379

    I just want an AI for infinite HDRIs!

  • CybersoxCybersox Posts: 9,053

    So, basically Stickychains StickyChain Plugin | Daz 3D combined with Kawecki's Ropemaker and Strapmaker products over at Rendo and/or AM's FoilageMaker The FoliageMaker Merchant Resource License | Daz 3D ?  Seems like the ability to do that already exists without the need for an outside AI product and, given that most AI engines can't even get basic things like hands right, I suspect the amount of cleanup work needed for to make an AI product completely believeable would probably require as much effort as using SitckyChains and basic rope products.  (Not to mention there'd be less worry about the eventual copyright issues that AI seems likely to run into in the near future.)  

     

  • DripDrip Posts: 1,191

    Some tool that would automagically fix little mistakes, like sunglasses poking through ears, or pokethroughs on clothes would be nice. But I just don't trust AI with that yet, I can already see it adding another sleeve to a jacket or something like that.

  • CybersoxCybersox Posts: 9,053

    Drip said:

    Some tool that would automagically fix little mistakes, like sunglasses poking through ears, or pokethroughs on clothes would be nice. But I just don't trust AI with that yet, I can already see it adding another sleeve to a jacket or something like that.

    Yeah, and with the current level of control offered in Photoshop and some of the other photomanip programs, there's very little that can't be fixxed in just a few minutes with a little postwork.    

  • FSMCDesignsFSMCDesigns Posts: 12,754

    Drip said:

    Some tool that would automagically fix little mistakes, like sunglasses poking through ears, or pokethroughs on clothes would be nice. But I just don't trust AI with that yet, I can already see it adding another sleeve to a jacket or something like that.

    yeah,  AI would case more small issues like this than fixing the ones you already have

     

    jdavison67 said:

    Phone cords! Ropes! Chains! Vines! Tubing!

    Imagine if you could designate a starting point, a rough path, and an end point, and let Ai fill in the rest.

    Think of it like Tweening in animation. you set key frames, and let the in-between frames get posed for you.

     

    JD

    Trust me, as someone that has been spending a TON of time with AI these past few months, you are better off using 3D products for this, unless you have the time to keep roling the dice (hitting the generate button) hoping for a great AI outcome. AI can do some amazing things, but only if it is trained to do so. It has no logic or common sense, only the data it is given and acessing that data with the right prompt is a pain at times. It's like if DS was run my AI with no content library or smart conetent and you wanted it to load a "petite", or "thin" figure, yet using those words as prompts don't give the results you want because the AI only knows "lithe" or "diminutive", so unless you use those exact words, you do not get the results you want..

    i was doing a crime scene image the other day where the victim was killed with a chain and it had to be draped across the body. I have a dforce 3d product for this, but wanted to see what AI could give me. I got great results (after htting the generate button over 100 times, LOL). So more realistic results, but more time to get them.

  • Oso3DOso3D Posts: 15,009

    What I'm really eager for is AI to punch up renders. Basically, intelligent postwork.

    I think there will always be a role for some degree of 3d scene creation, if only to place props/figures, pose, and do some composition. The more you can define and set exactly what you want, the better. And then sprinkle some AI magic to improve skin, hair, lighting, etc.

    Like, right now, Photoshop's generative backgrounds could be really cool as they improve.

     

    The big problem, of course, is that just about all the available solutions are wildly unethical. But that is a solvable problem. Here's hoping.

     

  • CybersoxCybersox Posts: 9,053

    FSMCDesigns said:

    Drip said:

    Some tool that would automagically fix little mistakes, like sunglasses poking through ears, or pokethroughs on clothes would be nice. But I just don't trust AI with that yet, I can already see it adding another sleeve to a jacket or something like that.

    yeah,  AI would case more small issues like this than fixing the ones you already have

     

    jdavison67 said:

    Phone cords! Ropes! Chains! Vines! Tubing!

    Imagine if you could designate a starting point, a rough path, and an end point, and let Ai fill in the rest.

    Think of it like Tweening in animation. you set key frames, and let the in-between frames get posed for you.

     

    JD

    i was doing a crime scene image the other day where the victim was killed with a chain and it had to be draped across the body. I have a dforce 3d product for this, but wanted to see what AI could give me. I got great results (after htting the generate button over 100 times, LOL). So more realistic results, but more time to get them.

    I think the most overlooked solution for this kind of thing is Photoshop's puppet warp tool.  I find that it's usually a lot faster to just use the basic posing tools, dforce, or whatever to get roughly there, render that one element out on a seperate layer, and then tweak the results in post by moving the little white dots around with the mouse. I remember the first time I saw Brian Habberlin demonstrate this trick at one of his ComicCon 3D art panels and being blown away at how simple and efficient it is for ropes, chains, long flowing hair, etc.    

  • jdavison67jdavison67 Posts: 645
    edited June 30

    Cybersox siad:

    I think the most overlooked solution for this kind of thing is Photoshop's puppet warp tool.  I find that it's usually a lot faster to just use the basic posing tools, dforce, or whatever to get roughly there, render that one element out on a seperate layer, and then tweak the results in post by moving the little white dots around with the mouse. I remember the first time I saw Brian Habberlin demonstrate this trick at one of his ComicCon 3D art panels and being blown away at how simple and efficient it is for ropes, chains, long flowing hair, etc.    

     

    OK... This jumped out at me as a possible road to what I need.....I must investigate!

     

    JD

    Post edited by jdavison67 on
  • ANGELREAPER1972ANGELREAPER1972 Posts: 4,505

    Tomhip said:

    I just want an AI for infinite HDRIs!

    you that's what I think it would be good for too, ruined building cities, overgrown, wrecked vehicles - aircraft/casrs/militery extra, alien worlds, underwater worlds, backgrounds/foregrounds objects that are too resource intensive or hard to put together or there are no products of for one reason or another

Sign In or Register to comment.