Adding to Cart…
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.You currently have no notifications.
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
I'm thinking I should focus more on terrain, architectural stuff, metals, and skin, for now, since the path forward on those is a lot more clear-cut.
For skin and a few others, I need to simplify the three axes of tiling to a simple 'scale' value; 99% of the time nobody is going to need to stretch in different directions, particularly with skin.
With freckles and veins and whatnot, I should add function to the actual shader, rather than trying to manage it more modularly with a separate geoshell layer (again, this confuses people).
So, skin color 1, skin color 2, freckle color, skin scale, freckle scale, freckle scatter distance, freckle fade.
Then maybe the same with veins (though that might be more for alien stuff).
I also need to look at other examples of skin and see if I can't do some stronger, more compelling effects.
Hmm. One kink I find in trying to do more 'smart material' style skin... so there's an effect, corners, where crevices in an object have a different color or pattern or whatever. This is a great way to enhance the realism and create some natural variation in an object.
I have that in WTP3 as 'grunge/corners verdigris' which is a great launching point to adding a layer that has this effect. Dirty corners, worn corners, etc.
The problem is, for figures? As the figure poses and bends and shifts, where this effect appears will also change; it's not 'fixed' based on the geometry, and it can't be.
Which mostly rules it out as a good option for most figures. Ah well.
I'll still try to do it for other stuff, like vehicles and whatnot.
(Specifically, I'm looking to create some presets that encapsulate these effects all in one shader, so people don't have to fiddle with geoshells. Geoshells are more versatile, but at the cost of set up effort and a bigger strain on the render)
Oh, and if anyone wants to do an 'all in one' for freckled human skin, you can easily do it by playing with the human skin presets, making the second color more distinct, and playing around with the Base Color values.
(IE: Set threshold values closer together, so there are identifiable blobs of color, adjust scale, though the default should be close to right)
Also, remember you can always turn on top coat; I have it off by default, but .1 top coat weight, maybe change the roughness to .3, and you get a nice slightly sweaty look.
Thanks to a number of helpers (HPhoenix in particular), got a breakthrough on the brick tilers so they can be rotated easily.
This is important for trying to match architecture with the shader; previously I'd run into instances where blocks would be obviously laid the wrong way (like, vertical), and there'd be nothing I could really do about it.
It is so good to see you back at it again Will!!!
Working at a 'simple brick' pattern, and aiming it at one of the most difficult challenges I have -- Jack Tomalin architecture. Even with almost every image map removed and ghost lights, it ran like a limping dog.
That would be difficult to judge because of the huge amount of emitters and the windows. Try the comparison without the window textures and don't use the mesh lights.
Connatic: Well, the point is whether using procedural shaders to replace where they can is useful, and in this case the answer seems to be 'yes.'
Another thing I'm working on is specific shaders useful for other purposes.
Like, right now I'm working on a stripped down Iray shader that eliminates a lot of options to, hopefully, make for faster renders. I'm motivated because a lot of people talk about how slow Iray is, but they are essentially comparing surfaces with loads of options to a 3DL render with very limited specs. When you actually up the realism of 3DL to be similar to Iray defaults, 3DL runs about as fast as Iray does in CPU mode.
So, my hope is that by using stripped-down shaders, the surfaces will render faster, and Iray will at least be more competitive with 3DL for 'simpler/fast'. Iray will never be QUITE as fast as 3DL, because the light calculations are inherently more complex than the simplest 3DL options, but still.
Sorry, I misunderstood. My main reason to use procedurals or any alternative is to reduce the textures' V-RAM consumption so my card can do the render. It will be fast if the gtx1080 does the render. otherwise cpu will be much slower. This is what is most important.
Exactly, but sometimes the 'cost' of removing stuff is too high. For the windows, the 'shape' of the window panes is defined by an image map, so I pretty much have to keep it.
However, I've done some testing and I realize that the benefit is REALLY situational. I made a regular Iray shader, but stripped down with most of the bells and whistles removed. And I found that it ran at 1/3 the speed of equivalent regular Iray shader. ... Argh.
So it's a ... fiddly, situational advantage. It seems like using Shader Mixer 'costs' a 1/3 speed reduction (at BEST) over optimized native shader. BUT... if you are facing a scene where the VRAM consumption is going to dump you to CPU, and CPU is less than 1/3 as fast as GPU (which it probably is), then it's worth applying new shaders.
So, basically, if your scene can run in GPU, great! Don't bother with special shaders. If it can't, first 'blank' the shaders of everything not visible (like the inside of everyone's mouth that isn't visible). Then simplify more distant stuff (like figures at medium-long range, replace skin with a simple color with no translucency and thin shell on)
If that STILL doesn't fit, start replacing with procedural stuff.
The other two situations to use procedural shaders:
Distant/large shots where huge sweeps of terrain/architecture will look obviously tiled.
Close shots where the resolution of image maps becomes obviously pixelated.
Out of curiosity how many CPU cores / threads and GPU did you get earlier this year Will?
I have an Alienware Area-51 (R2)
Intel i7-5820K CPU @3.3 GHz (12 CPUs), 16 gb RAM
NVIDIA GeForce GTX 970 x2, which have 1664 CUDA cores each (the first card was from my previous computer whose harddrive failed, so I had elected to get a second and pair them up).
VRAM is limited to ~3.5 GB, which is definitely a pinch I can feel on the bigger ones... but so long as the scene fits, it then runs pretty fast. Extra incentive to develop ways to reduce texture load. ;)
Thanks.
Wooo...with those specs I'm a bit disappointed to hear you complain about render times of anything so now I know the slowness you speak of must be due to the 12 threads rendering procedurals? Just asking while I'm considering a 16 thread Ryzen later.
Just what procedural can the GPUs render and which, if any, must be done by the CPUs?
It's not a matter of it being slow, it's a matter of it being slowER. And you can always push it. Jack Tomalin environments/buildings really push my machine to the limits.
All of my shaders run slower than Iray Uber shader, because the Iray Uber shader is engine optimized. That SAID:
Pros of Procedural shaders:
Procedural shaders have no texture load. If you are right at the edge of memory dumping you to CPU, a few procedural shaders can pull you back.
Procedural shaders mostly lack the scale problems of image maps, so look good at long range and short range. At huge extremes you might see some tiling or pixelization, but at much greater extremes than image maps.
Interesting tweakable patterns. Like, you can have freckled skin. If you want the freckles to be a different color or further apart or bigger, trivial to do that (unlike image maps).
Many procedural shaders ignore UV, which has several knock on benefits: you can apply the same shader to different parts of a conjoined figure or geograft and it all flows together; the object can lack a UV map; the shader will not get stretched out due to mesh distortion.
Cons:
Slower, because Shader Mixer isn't as optimized as in-engine shader.
Mostly homogeneous. This is particularly a problem for skin.
A little more fiddly. Among other things, the user may need to adjust scale for different types of objects.
OK, I'm used to slow so no biggy for me.
So I'm FINALLY releasing my bubblehead and tentaclehead stuff, plus a slew of weird skins (orca, metalloid, fleshling, fungal, pebblesquid), and there's a lot of stuff for almost everyone. Like, fleshling materials for octolegs and four arms! HAHAHAH!
The sucker is also 1.55 GB, so... um. Hmm. Going to be a while
(Having to break it up into chunks since sharecg is like haha ha no)
Cool, I need to take a look. I'll have to stop my iTunes downloads for a while. LOL.
It's probably going to be another 30+ minutes before I get it all sorted. ;)
Ok, uploads FINALLY done: https://www.sharecg.com/v/87890/view/21/DAZ-Studio/Head-geografts-and-skins-Genesis-p-1-4
cool thanks
So I'm contemplating trying my hand at retexturing old content.
Out of the following, which would you most like to see new textures for?
https://www.daz3d.com/tawhak
https://www.daz3d.com/slime-beast
https://www.daz3d.com/plant-pixie
https://www.daz3d.com/millennium-cat (but only for something weird, like slime cats or scaley demon cats)
https://www.daz3d.com/millennium-dog-le (ditto)
other?
Note that anything with hair is incredibly difficult, particularly to get it to flow and fall along a surface. So, starting out... no hair.
Of thos I have only Mil Dog LE, I don't know if it has a chihauhau morph but one of the weird hairless dogs would be interesting.
I was thinking of doing an alien dog at some point. Hrm.
Sounds interesting
Interestingly, some of the older stuff is actually easier to work with, because more often most of the body is one surface, which gets around some of the issues Substance Painter2 has.
I've only ever created a new set of surfaces for the DAZ Seal Hi-Res is you want that preset. Sort of a doggish look to them.
Took a quick wack at the slimebeast... pretty happy. Didn't end up replicating the eyes (couldn't do them quite right and got impatient).
I have a bunch of ideas for dog skins, though, and maybe even some morphs...
yuck, that's pretty convincing slime
You know, I was grousing about the limitations of 'no HD morphs' and model sculpting Daz imposes, but I'm realizing I don't think I have a powerful enough machine to handle high end models anyway, and I definitely don't have the best tools for that.
So, hey, Base morph + texture seems to be the ticket...
So, short term goals:
2-3 skins for Slimebeast (for the few people who own that)
2-3 weird Daz Horse 2 skins (already have Scalehorse done, may touch it up. Thinking of maybe a rocky magma horse? An Iron Golem horse shouldn't be too hard. Mmm)
A few skins for the Martian riding beast. Delightfully alien looking already, so...
All free, of course.