Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
I used to do a lot of skin shaders for other PAs, it took me sometimes a couple of days for one characters.. I pay my rent and feed my family on this, please don't say good skin shaders aren't valuable.
That's a good one. Optitex on DS is very interesting. Not to say, it involves some software for making cloth 'Patrons' to feed to a sowing machine factory system (for making actual cloths). The software in question is a yearly rental like PhotoShop in a way, and it only supports a single sowing machine format natively (without an adaptor). To use other sowing machine types, you must use a web based service to convert them (at the time it was a pay per single use setup), and OBJ/DSF/DUF is not in that list of 'suported sowing machine formats'. The converter to export the Cloth patron from the Optitex suite software to DS is only available from a third party, that just so happens to no longer work for Optitex. I'm thinking this is starting to sound rather, uh, lets just say I don't want to know what's going on. *looks around to see if the police are near* because whatever it is, it can not be good, lol.
The Dynamic tab module software stuff in Studio also has some difficulties as well. I was able to get around it's biggest issue using the "G-Suit 2 HD" on G2F as a collision shell to keep the outfits out of the figure. It also has little in the way for setting up force directions, so most things must be animated prior to the single snapshot render. And many things are only for generation 4, and require starting at T-pose at world center and walking your figure out to the finale location and pose for the Dynamic stuff to not turn into a ball or just fall threw the floor.
So, if your still into V4 stuff, and you are OK with Photo shopping your renders to make it look like nothing is wrong. There are a few very nice threads around with lots of advice on using the Optitex outfits sold around the net.
(Links Pending)
Using V4 Dynamic Outfits on G2F...
http://www.daz3d.com/forums/discussion/comment/737036/#Comment_737036
Me and Kettu did apologies, as that was Not implied. Shader presets are incredibly valuable to many, and do help save time and effort when setting up scenes. It is very nice when you don't need to set up everything in a scene from scratch. Much the same way that light presets are also very nice to have.
So, let me wake up to some coffee. I think this is a my muse left me day, lol.
I read everything..
(walking on Glass) From what I gather, Subsurface works by calculating how light passes threw the surface. 'Scatter' is with the light moving threw the mesh to illuminate neighboring areas on the outside normal, and I think 'Absorb' has something to do with the amount of light coming out the other side of the surface (illumination from the back side?).
This is calculated for each Poly (or quad) on the surface, so the more polygons, the more calculations are needed. also, when the shading rat gets involved, it also controls how many times the scatter and absorb is calculated. And to top it off, it appears that the number of subD poly's also impacts how intense this calculation gets.
So, non-hd figures will have a shorter Sub Surface Precompute delay, then HD figures. And pleas don't read that as I prefer non-HD figures, I like them all. Otherwise I would not go threw the trouble of making AltShaders to make life easier using them in scenes.
I know it can get much more involved then that. Tho before I finish my first cup of coffee, this is what little I think I grasp at the moment.
Sounds about right... my way of making 3DL shaders is 1) Does it look plausibly realistic 2) Does it not take OMG forever to render 3) Try to find reasonable balance between those two. I really could care less about that "real world" values are supposed to be, about the only one I ever pay attention to is index of refraction.. the rest I just eyeball it, 'cause reallt it's not gonna matter in the slightest when a 10% stronger light makes it look completely different. And I'll usually go way lower on the strength of SSS and translucence because those effects are super picky about lighting.. setting them high enough to look really creamy/waxy in the light I'm using to develop might (likely) end up with 1/2 the end users rendering orange sherbert characters.
Yes, and it helps to think of "absorption" as "inverted transmission". That is, if the absoption is blueish, the light that goes out will be orange-ish. It's not as straightforward with SSS as it is with "simple" absorption for glass-like materials (Beer's law) because the light doesn't only pass through but also bounces inside somewhat, but you can guess the overall tint from the "transmission" colour anyway.
With AoA SSS, though, it's difficult to visualise those scatter/absorption colours because the shader mixer SSS brick uses sliders only and not a combo of colour swatches + depth sliders.
Especially free presets with extensive documentation, such as those of mine =P
I've noticed some talking about stuff like that in the Iray skin discussion thread. All kinds of interesting effects when lights play with surface settings. I'm not all that fond of getting in to the time consuming process of setting up lights that only illuminate one thing (Specular lights, Diffuse lights, etc). To me, light s light, and it is enough to fuss with light angles for a particular pose and scene.
Now I'm not sure if there is a penalty of sorts to render times with having lights from a different shader set from the surface shader, tho I would not be surprised if that also contributes some to Sub Surface Precompute delay and the render time. It may also be that it maters not at all, as light is light in 3DL regardless of what shader produced it.
Kettu, I'm not sure how I managed to miss those lights. I'll look around there in a second. I have gears spinning in my head, as I'm connecting some dot's.
From what I can tell The Uber Environment light takes longer to render with the AoA surface shader that the AoA Ambient light does.
Doesn't AoA Ambient do some internal optimisations specifically for AoA Subsurface?
Besides that... In another thread, we suspected UE2 to be doing some very unusual things under the hood (more like the original Pixar Renderman style of coding than the 3Delight-specific style), which often lead to longer render times as compared to more straightforward 3Delight-only light shaders. But it's the only environment light readily and widely available for DS that supports image-based lighting. So there's a tradeoff here as well.
Yeah I believe so.. I alway used the Uber light anyway and ate the extra render time 'cause I seemed to get better end results out of it. Now i just use Iray unless I making matfiles for a product
Yea, and I have a vague grasp of that collection of settings in the AoA Sub Surface section, and the numbers in the dials may as well be in Greek, lol. I have no idea how the Scatter and Absorb combine to tint the map in the Sub Surface channel, and I just don't have the patience to spin the dials and wait for the spot render to show me the results (Especial when I'm already running three different directions, lol).
So. How dose that all work into the render time of a figure, it also matters because of what the SubD is doing with the Sub Surface Precompute delay. Now I'm guessing that HD figures have finer details in some locations, or just everywhere for that particular figure shape and the details. The dial in the Parameters tab is only a lower limit, and a figure shape can do more subD then what the dial is set to. I just looked at three different random HD figures, and they all have the Dials set to SubD 1 in the view field, and SubD 2 for renders, Brook is also using the same setting there.
I'm guessing most of Brook's details are in the Displacement/Normal/Bump maps, rather then using more SubD. That matters for the explanation, for my former question. The details are in the maps, not SubD mesh. Sub Surface Precompute is calculated for each Mesh poly, not the map pixels. Map pixels are added to that after the render begins. So the more a figure relies on SubD for the details, the harder it will be hit with longer Sub Surface Precompute delay (Face Plant Time).
Now I'm guessing that if I convert Brooks mats from AoA over to Omni, it will still render insanely fast, because most of the details are in the maps. I may try that some day, not today. There is a few other things I need to look at, on other topics and things.
That gives me a brilliant idea for the 3DL lab (evil laugh) Where is the AoA and Omni "Lazer" (evil laugh). lol.
http://www.daz3d.com/forums/discussion/55128/3delight-laboratory-thread-tips-questions-experiments/p1
No Mr 3DL, I don't expect you to talk, I expect you to render. lol.
Has it been timed the other way as well? AoA lights on Omni surfaces. And AoA on AoA, then Omni on Omni. Hmm. I'll need to set up some Spheres cones and cubes. Not to say, find the aforementioned AoA lights. I know where the Omni lights are, I'll need to look to see just what I do have for AoA stuffing (oh, wrong season.).
And I wrote a completely custom suite of 3Delight shaders and render scripts that take advantage of its newest features and do exactly what I want them to do =)
Wowie seemed to have some tests here, but they were on an older version, and the farther 3DL develops, the slower UE2 seems to render =), especially in GI modes.
http://www.daz3d.com/forums/discussion/53148/aoa-lights-vs-uberlights-vs-daz-standard-lighting/p1
What I know, though, is that I am unfortunately not interested in measuring/optimising the interaction of someone else's shaders anymore =) So if you're up to the task, you'll be a hero of those who still use the stuff.
3Delight technically always renders the so-called "limit surface" of a SubD mesh, so the level setting shouldn't affect performance. On the other hand, remember Mjc's recent posts? The way DS passes the Catmark SubD surfaces to 3Delight is sort of mysterious and makes the render slower. And slower not only for the SSS pre-pass, but slower in general. This seems to go away when the scene is exported to RIB and rendered in the standalone.
(Core dump) The universe hates me. This was supposed to be a nice simple day of spinning dials and looking at settings.
I know I set that thing to NEVER do this, and now I have at least a few problems because it decided to just do it any way. I was in the middle of doing a Spot render, and Windows7 just decide to spontaneously Reboot for a Windows update, in the middle of the render. No warning, No pop up message saying the comp needs to update, just instant reboot into the update screen. WTH, I know I disabled that, months ago.
Grrrrrrrrr. lol.
Sorry about that. Windows can get annoying, as if on purpose...
It's some super duper, extra special, if you don't do it now, there will be snow in Alaska, Hell will freeze over and the political parties will actually throw a party, security update that MS has overridden everyone's settings to force it to happen...
Measuring for the sake of knowing what's going on is one thing. Fixing or adjusting some one else's stuff, na, that's bad, very bad, for many reasons.
I've looked over the settings from a few angles, and it's the only explanation that fits. Brook has a Face Plant under forty five seconds, and every one else is over fifteen minutes. It's not a more SSS then Diffuse (there all doing it), it's not something turned off (they all have Diffuse, Gloss 1 & 2, Velvet, and costume SSS), and there all G3F figures at a SubD of 2 in render.
I do remember that note about the RIB being faster for some stuff, tho I have yet to get a second comp running with the Stand alone 3DL on it (limited funds). It would be very nice for a few reasons. Not to say, I could have it rendering something while I mess with another scene at the same time, instead of sitting here twitting my thumbs waiting on the render, lol.
It's all good. I got Studio's interface back to normal, and it doesn't look like the database got scrambled (knocking on wood). I'll deal with the corrupted HDD later, when I get tired of not having music.
In the past I use to get a window that let me know there was an update, and I had a chance to save and close things down to do the update. This was literally just a spontaneous reboot without any warning. So I need to find out if the settings for that got wiped or something.
(EDIT)
lol, that sounds about corect MJC1016, lol.
Music, something the "Raspberry Pi 2" is almost capable of, lol.
You know, that very well may be. It was back on the Don't warn me setting. ha, odd. It may have been a previous update that wiped those settings.
Our HTPC did it earlier this morning. Had to log back in so my daughter could finish watching her show.
So I'm thinking of a few tests, that I would like to look into, and comparing AoA to Omni Sub Surface Precompute delay directly may be a tad difficult, as they both do things quite differently, and that is not exactly what I want to look at. The bigger question is how SubD effects render times. So I guess this will require looking at how increasing the SubD on each of them effects render times. I think to save my sanity, I will be doing this on a simple cube and comparing times, tho It will take a bit of time to figure out just what settings I should use on each shader for the test (Sub Surface settings are still a hazy area for me). That and I'll head the opinion of Daz_Spooky, and yes, the title of the thread gives me considerable concern for those of us on mere Personal Computers.
http://www.daz3d.com/forums/discussion/comment/1011688/#Comment_1011688
I am not taking Daz_Spooky out of context here, just noting that I have experienced what is pointed out in this one sentence from that other thread. There was a brief moment in time, where some figures had independent HD dials. It was necessary for some of them to turn the HD dial off while setting up stuff the figure was interacting with. In a way, I regret not having those dials today, and at the same time I understand it is another thing that takes time and effort that most PAs don't have to make and set up.
So, while I let the gears turn on how best to do the tests and compare the results, I have a little fun experiment I'm starting to set up. Here is a screen-cap.
Left is EJ Estela, Right is FWSA Yulia. I've puled them into the scene from the former Hair test scene, and started setting up some stuff (still in progress). Estela on the left is wearing the 'W Skirt' over the Bunny outfit (that may change). Yulia on the right is wearing the new Midnight Club outfit. Now I just tossed a random pose on Yulia, and I'm going to do something a tad different this time. In the past, I would take such ridiculous poses, and adjust the joints back to something a tad more reasonable for the range of human joints, and the skirts over them. This time, I'll reluctantly adjust the outfit to not be so kinked up. There is one minor adjustment to the pose, I moved the right leg back and in a tad to get the feet back on the ground, as the pose was not set up for high heals.
I guess it is well known that I'm not a big fan of high heals, if anything, I tolerate them, lol. I will say these ones do look good. As for the rest, well That's what I'm going to work on next.
Where is Tenshi, and the "Midnight Club Nightlife Textures" from the "Tenshi HD and Midnight Club Bundle". Well, honestly, there in the "Will buy" list along with the EJ "Wild Side Outfit". I spent this month's play funds on hair styles, that I hope to try out on these Elves.
Well, the two glaring things in that first screen-cap, is the bracelets, and the shirt tale. When I first put the outfit on Yulia, I went threw the stuff that was right there in the scene tab, and put mats on them. The bracelets however are not down on the root node, as there not conforming items. Instead there like quite a few other things I have that are parented to the bones where they are located. I find this to be a lot easier to adjust the location of them for any particular pose, where a lot of conforming stuff is a tad more tedious. I have found in the past that some conforming things have the translate dials disabled, making it impossible to use them. So in any case, I'll select them and put some mats on the bracelets. Adjusting there location on the arms is as easy as moving anything else around in studio.
So, got some color on the Bracelets, and used the arrow things in the view field to place them on the arms.
Now the shirt tale is going to be a tad more interesting. It has quite a few adjustments for fixing poke threw, and a very nice setup for bones.
As mentioned before, all I did so far is toss a pose preset on Yulia.
Before I start playing with the Pose controls, this Keyboard is Fired. I have no idea what layout has the backslash where the back space is supposed to be, tho that is NOT Qwerty, not for the past thirty years at least. I gave it a try, Now I'm going back to the Original Saitek that worked for me.
Going by the 3Delight docs AND the tests I did...it's not really the subdivision that's the thing that chews up the time...it's the subdivision algorithm. It seems that part of the precompute process is translating the OpenSubD algorithm that the HD morphs rely on into the 'legacy' Catmull-Clark that 3Delight uses.
There is something in the way Studio processes the info for the render causing it...and there is a correlation between the algorithms and the time. Catmark is measurably the longer one.
Bilinear and Loop seem to be written out to the RIB as mesh data, not a 'tag' saying 'subdivide this simple mesh'. They both render fairly fast. Catmull-Clark, on the other hand, is simply a line saying 'subdivide this mesh'. And it flies...both from the smaller RIB that is generated and the fact that the subdivision is only taking place as and when needed.
From the 3DL docs...
Catmark, it seems, when sent to a RIB is treated as Catmull-Clark...but could it be when rendered from within Studio it is being written out as not a 'subd' mesh, but one, that first is being fully tesselated and then sent to the included 3DL? Doing that WILL eat up some time...and probably with a higher level of SubD (as in HD morphs) more time to do the tesselation and send it off.
We really need one of the devs to comment on that to be sure...but that's what it seems to be doing. Another piece of supporting evidence...it takes longer to write out a Bilinear or Loop RIB than it does for either Catmull-Clark or Catmark (and they are MUCH larger RIB files, too).
One thing from what I just did...I discovered that Bilinear is best at preserving edges...even at insane levels of SubD, a cube looked like a cube...not something that morphed into a sphere.
MJC1016, That is a very interesting thought. Tho I'm not sure it has any Bering on the Sub Surface Precompute delay, at all. Both OziChick Boork (less then 45 seconds), and "Face-plant-goddess" (bloody ages) are both using 'Catmark' by default. The other G3F HD figures, I can't remember if it was around fifteen minutes or closer to half an hour for each one, tho they also use 'Catmark'. It dose bring up an interesting idea for those of us that only have Studio without that 'Stand alone' 3delight stuff.
As for changing that to another setting, I'm not sure if that would have any improvement for G3F clothing Smoothing at all. Not that it would hurt the existing G3F swimwear on non-daz G3F figures, lol.
The OziChick settings could be related...all I know that a simple cube with AoA on it and SSS turned ON takes a bit longer to have the first pixel show up with Catmark than it does with Catmull-Clark (progressive ON).
I just ran a test using a simple six poly cube subd level 7, DaWaterRat's Exalted Shaders (they are AoA presets: http://www.sharecg.com/v/71898/gallery/21/DAZ-Studio/Exalted-Magical-Materials-Shaders-Updated) with a single 'standard' spot light, with raytraced shadows.
Catmark: 55 seconds to first pixel.
Catmull-Clark: 52 seconds.
No, not much but it was consistent over several runs...always about a 3 second difference.
Dropping the SubD level back quite a bit...(to a more normal '3'), with the same light...and now it was 18 seconds for Catmull-Clark and 21 for Catmark!
With a more complex mesh and more surfaces, I would expect the gap to be wider.
I got sucked into watching a show, oops. So, it didn't take long to tos a few of them posing dials and find a mix that worked for this. A little bit of a 'step', 'float', and 'Flare Right'. Then I spent the rest of the evening looking at different color combos for the stuff EJ Estela.
That is Tsukiakari Hair with an included preset Auto fitted onto Estela. The octagon shaped Bracelets are from EJ Tatijana and Boho Jewls. That W-skirt came with that color preset, and the stockings were done up using the AElflaed's Fancy - Shaders. The Stockings and high heals are from the Bunny Girl Costume. The back drop is part of Stonemason's Pool House, with just my test chamber lights. Now that that is out of the way, on to Yulia.
As good as this looks, the shirt tail on her left side, can use a little mor work. Luckily there is a few bones to allow doing some really cool stuff.
And while I'm messing with bones, I'll tweak the right side as well.
And here is the result.
At least the lack of HD, tho I don't know how to find out exactly what the shaping dial is doing to the upper SubD limit. It would be interesting to know just how far the SubD is being raised by particular figures, and how that relates (if it dose) to the Sub Surface Precompute delay. It would also give me an idea if changing the shading rate settings in the AoA shader would help or not (assuming that setting dose anything at all for Sub Surface).
Thanks for the link, I'll look at that. I am curious about that one note on how 3delight deals with SubD, especially for a simple cube that dose not need to be more then six quads to make it. Something to look into after some rest.
(EDIT) yea, EJ Estela, FWSA Yulia HD, and OziChick Brook, are all showing the same exact values in that 'notes' thing in the scene tab.
Vertices 17,418 / 68,744
Faces 17,000, 68,000
All three are very different as far as root shaping. Estela is a dial-spun body with a master control dial and a custom head/face (I don't know if any of it is 'HD' or not). Yulia is a custom head&body shape, and absolutely HD. OziChick Brook (attached screen-cap) is a custom head&body shape, and to the best of my knowledge is not HD.
Everything else being the same, it's the same as with any other shading rates: lower values = more render time AND higher quality. Generally, how high you can go will depend on the SSS scale: higher scales can tolerate very high shading rates w/o artefacts. But I personally think skin looks better at low scales... but then everyone knows I primarily render ethereal faeries, solar avatars and other nonhumans =)