A new UHD line of characters with 800'000+ vertices at base resolution?
linvanchene
Posts: 1,386
Another thread asked the question what is needed to create more realistic images.
Because that thread focuses on what can be done with the assets available now I split this up to a 2nd thread because it points at areas out of the reach of the customers.
To give the DAZ3D community a true leap in quality there could be a character generation with
Higher base model resolution
Have a quick look at
http://www.zbrushcentral.com/zbc-top-row-gallery.php
- - -
To paint and model fine details you may want a resolution of at least 1'000'000 vertices.
To go even further with HD geometry you may want to work in some areas with 50'000'000 or more vertices.
So how do you go from a model with 100'000 vertices to a model with 1'000'000 and more?
Subdivision levels:
When you add subdivisions each polygon on a model gets split into more vertices.
The original basic subdivision stays exactly the same.
But now at the higher subdivision levels you can sculpt and paint details.
Instead of having the whole model at an insane high vertices point some modeling applications allow you to add more vertices in just one area.
Example:
http://pixologic.com/zclassroom/lesson/hd-geometry
- - -
How do you bring the details back to DAZ Studio?
Currently for those with a modeling application there are four options to create high detail and bring it back to a DAZ3D figure each with their own advantages and disadvantages:
- Export / import model at higher subdivision level
If your system can handle it you can just export an .obj with a vertices count at 1'000'000 or even up to 25'000'000.
The advantage is that you can create details that can not be captured by maps or morphs.
You can alter the original mesh in any way you want.
Example:
Very high detail Hair, spikes and other curved forms that are not just sticking out in the normal direction of the model but end up in spirals or other more complex forms.
- Displacement maps
The advantage of displacement maps is that you can have the model at the base resolution and then apply displacement maps that capture all the detail.
You can alter the higher subdivision levels in any way you want
The downside of displacement maps is that unfortunately many GPU render engines do not handle them that well currently especially on lower resolution models.
- Morphs
The advantage of morphs is that you can dial in the details at different levels.
The downside of morphs is that you cannot alter the original subdivision levels.
This means all the details you add need to be created by moving the vertices around to other positions.
You can create some spiral shaped forms that you cannot create with displacement maps.
- HD morphs
The idea of HD morphs is to add details on a higher subdivision level.
This has the same limitation as normal morphs in so far that you can only move around vertices.
But because you can sculpt at a higher resolution around 1'000'000 vertices you can create some finer details.
The use of a tool to transfer HD morphs back to DAZ Studio is currently limited to official DAZ3D partners.
- - -
One type of figure with one base resolution is not able to fulfill the needs of different user groups
To summarize in a very simplified form
If you want to create more realistic images you need higher resolution models that are able to capture all the details.
Only if the actual original base geometry of the model has more vertices you can create all the details you want without being limited in one way or the other.
From that point of view a base figure created for high detail sculpting should have a vertices count of around 800'000 to 1'000'000.
[Start of Edit]
Updated / Edited to be more precise:
I removed a link to a 3rd party blog comparing different generations of Victoria and their vertices count because of user input that those numbers may not be accurate.
When you compare base and high resolution levels of Victoria 4, Genesis, Genesis 2, Genesis 3 you will notice that they stayed pretty much around the same level of vertices. In fact the from Genesis Generation 2 to Generation 3 the vertices count was actually lowered instead of raised.
When I tested this myself I found that G3F has a base resolution of 17'418 vertices and a high resolution of 68'744.
What is called "Base Resolution" in DAZ Studio is called subdivision level 1 in Zbrush.
What is called "High Resolution" in DAZ Studio is called subdivision level 2 in Zbrush.
[/End of Edit]
- - -
To me it seems in the past the goal was always to provide models that could in theory also be used for games.
The vertices count may have been low to make sure old systems can handle it.
It seems that DAZ3D now has a side company specialized for low resolution game models:
Maybe now that the needs of the game developers are covered is there not also room to provide a more higher resolution model optimized for high detail sculpting?
- - -
3 different product lines of character models?
Maybe there would actually be room for 3 different types of character models by DAZ3D?
- Genesis for the average users
- Morph3D for game creators who need quality at lower resolutions that is optimized for real time rendering
- A UHD figure Generation for users who like to use DAZ3D characters and sculpt at very high resolution details
The underlying idea would be that at least clothing could be shared by all 3 product lines but character textures and morphs would be unique.
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
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
None of the Genesis models have as may vertices as you have listed, they all have roughly 20K. That only becomes around 80K after 1 level of subdivision.
Only Daz vendor have access to the HD morph plugin, so you may have to ask DAZ them self to get your hands on it. One of the perks of being a vendor here !
Displacement mapping is only workable with 3Delight as that is per pixel, Iray is per vertex !
I assume iRay displacement maps would work on a heavily subdivided model, no?
Yes that's the best way around it you set the sub-d level on the material I've never tried it but it should work 16bit tif I assume
I just posted these images of some morphs on the V4 basemesh in the "Show us Your Iray Renders" thread, but I'm going to post them here again because I'm proud of them. As Havos pointed out, The Genesis models have 1/4 as many polygons as you list since it seems you listed them after subdivision (set subD to 0 and mesh resolution to "base" to see the actual base mesh of genesis figures). The V4 model on the other hand... it has the perfect amount of polygons as far as I'm concerned. Enough that displacement mapping is not necessary, any further surface detail can be had via normal and bump mapping. The genesis models though require either displacement maps or HD morphs to get that level of realism.
An 800,000 poly model would be a bit unwieldly to pose on today's machines unfortunately. I can import an 800,000 poly .obj into DAZ and render it, but it is noticably taxing the system and that is not even trying to rig and fit clothes and poses on it. I use zbrush every day and love it, but I'm aware that it has to pull a few tricks to sculpt on 20,000,000 polys. Riggin a figure that detailed and fitting clothing to it in a program like DAZ isn't something that has been possible yet. In the future though, I agree with you - let's get as many polys as we can. Also, I think that the whole Genesis line has too few polys for my taste, and like I mentioned before V4 and M4 hit the sweet spot right on the head with around 70,000 poly base mesh.
Anyway, not that this compares with those hi-res sculpts you linked to on the zbrush gallery, but here is what I did tonight with V4's basemesh resolution and normal maps:
Not bad at all for V4 I can see the details on the shoulder, nice !
Thanks! Yeah, the silhouette of the shoulders on the Genesis line are one of it's major weaknesses. The shoulder bends beautifully well in Genesis 3, but seen in silhouette there are only like four or five polygons over the main curve of the shoulder's silhouette - not enough to get realistic anatomical detail. V4 bends relatively poorly out of the box, but with some commercially available add-on fixes and custom fixes of my own, the V4 and M4 figures are my current favorite. Here's a comparison of the unsubdivided basemeshes for V4 and V7. V4 has about 4 times as many polys.
Thank you for clearing this up. I updated the first post to put the quoted information into perspective.
I was googling for a post about a comparison of the last generation to get a rough estimate.
I noticed that there are some differences to my own findings but the rough estimate seemed about accurate.
The point was simply to illustrate that the vertices count did not raise in any way but was actually lowered from G2F to G3F.
Here are the numbers that I was able to confirm myself.
G3F
base resolution of 17'418 vertices:
high resolution of 68'744 vertices
What is called "Base Resolution" in DAZ Studio is called subdivision level 1 in Zbrush.
What is called "High Resolution" in DAZ Studio is called subdivision level 2 in Zbrush.
If you export G3F at the base resolution level of 17'418 vertices and subdivide it once you get to the same 68'744 vertices.
- - -
This means if you export G3F from DAZ Studio to Zbrush you are working with only 17'418 vertices.
This severly limits the detail level of default morphs that you can create.
- - -
Are 800'000 vertices really too much to handle for an up to date workstation in 2016?
Even with my older workstation in 2013 / 2014 I had often scenes of 25'000'000+ vertices.
With fiber mesh hair you reach that number rather quickly.
I do realize that many customers of DAZ3D may do this as a hobby. Therefore the Genesis line of products must stay at a reasonable vertices or polygon count.
That is why I suggested to create a new UHD line of DAZ3D models with larger BASE vertices count so those users who want to make use of sculpting workflows including 3rd party software like Zbrush have more options to create high detailed models.
If you are using dynamic clothing, then simulations run much slower if there is a large number of polygons. Naturally with well written software this could improve dramatically
A couple of minor points.
1. It isn't just GPU based renderers that require subdividing the mesh to make a microdisplacement map work, but virtually every render engine that is not Renderman based (Renderman based includes both 3Delight and Firefly).
2. Displacement maps are reliant on the UVs of the model.
3. Displacement Maps only displace along the Normals of the face displaced.
4. You have to be careful with the amount of polys in all render engines (Displacement created or subD levels) as that will increase your render time and amount of memory used. I knocked a 12GB K6000 out of my Iray render by setting a piece of armor (which had great scroll work on there in a displacement map) to Tesselation level 6. For the other side of the coin compare rendering an HD figure with clothes and hair, and use area lights in 3Delight, to the non-HD version. (See you sometime Sunday.)
More polys do not, necessarily, equal more realism, but they do equal more memory used and more render time. In many cases Normal Maps are sufficeint to create added realism, without adding a single poly.
There is no magic bullet, you have to be careful with what you use and how you use it.
Does explain why I rarely see the kind of morph details that I see in some V4 packages. Without HD morphing or displacement map trickery, you simply can't get the same level of morph detail. Turning off the subdivision on the genesis figures is kinda eye-opening when you realize how few polys they really have. Though at least it's something for me to keep in mind when using more than one in a scene as I can do without a lot of detail for background figures.
interesting!
what about carrara displacement, is it per pixel or vertex?
Thank you for adding a lot of points that also have to be considered.
Nevertheless I was always at the impression that with GPU rendering we have to use Subdivision level 3 or even 4.
Therefore there is no benefit at all during rendering to have a 17'418 vertices base resolution.
During rendering the calculations will be made based on the 1'090'700 vertices of the subdivised version.
The VRAM used in GPU rendering will be based on the subdivision that is selected.
- - -
I made some comparisons of VRAM usage:
- - -
G3F at high resolution: 68’744 vertices
VRAM usage:
Est. Textures : 516MB
Geometry & others : 291 MB
Total Est.: 807 MB
- - -
G3F at DAZ Studio subdivision level 3: 1’090’700 vertices
VRAM usage:
Est. Textures : 516MB
Geometry & others : 658 MB
Total Est.: 1’174 MB
- - -
G3F with 1’090’700 vertices at subdivision level 3 uses only 367 MB more in VRAM than a default G3F with 68’744 vertices.
To me it seems some people are saying that their system cannot handle 800'000+ vertices but in fact most of the systems actually can and did.
Everytime you are using subdivision level 3 at the final rendering stage you are using a 800'000+ vertices model.
- - -
My conclusion:
It seems it is required to use a high resolution model at rendering stage anyway so HD morphs and displacement are calculated properly.
Therefore we might as well use a high resolution base model and benefit from all the advantages that would offer when creating morphs or polypainting directly on the base model.
WOW.... some VERY impressive work - expressions are life-like.... Kudos!!!! - Ray
Unless you are rendering a facial close up, the HD facial morphs are almost impossible to see, and so add very little to the render. For characters at mid range and beyond there is little point using more than sub dev 1 (ie 80K polys for a Genesis character) and background characters can use base resolution. You would need to have incredibly detailed HD morphs to need sub dev 3 or 4, and those may get washed out due to details in the skin texture.
The question is at which subdivision levels were those HD morphs actually sculpted?
The next question then is how much detail is actually lost in the process of converting, transforming the subdivision levels from a 3rd party application to the DAZ Studio subdivision levels?
Unfortunately the product pages do not indicate at all if a HD morph was designed
- to be used at subdivision level 3 with 1’090’700 vertices
or
- to be used at subdivision level 4 with 4'357'308 vertices
- - -
A vertices count of 1’090’700 is just the mere basic starting point so you can see some detail when sculpting.
A reasonable sculpting mesh for close up renders would be about the next subdivision level that then jumps up to 4'357'308 vertices.
If the HD morphs were sculpted at 4'357'308 vertices then you also need to set the subdivision level to exactly the same amount during rendering or you will not be able to see the details.
If the HD morphs were sculpted at 4'357'308 vertices and you use a subdivided mesh with a resolution of only 1’090’700 you will NOT see all the detail that was originally sculpted.
For those interested the VRAM required at subdivision level 4 is:
4'357'308 vertices - 1888 MB VRAM
- - -
Agreed. It would be a waste of memory to use the UHD figure line as background objects. For this purpose the current Genesis generation of figures is perfectly fine.
Maybe I was not clear enough on that point:
I was not recommending to replace the current Genesis line of figures but to add a new UHD line of figures for those users who can manage the system and render requirements of a 800'000+ vertices base mesh.
The reason why I set the base mesh number at about 800'000+ vertices was because this is the recommended resolution for a base mesh for sculpting in ZBRUSH when you use HD subdivision levels up to 50'000'000+ vertices.
This is explained in more detail in this video:
http://pixologic.com/zclassroom/lesson/hd-geometry
- - -
I guess we will be stuck with the current limitations until someone is willing and able to provide a solution.
If someone wants to have truly high detailed characters for close up renders one has to create them in Zbrush and use vertices counts of 4'357'308+
Maybe in the future there will be some new technologies that solve all the issues and limitations of bringing all those details back to DAZ Studio.
But I do realize that when I am the only one interested there is not much of a market at the current point in time...
Very impressive; I'm all for more geometry.
Personally I think the hair should catch up with what we fave first.
If I could create expression morphs like that, I'd go back to V4 myself. You need to write a tutorial on how you create those morphs. Any chance you could create morphs like that in Blender?
I'm not a blender user, but I've seen that it has some sculpting functions. You should be able to make morphs with blender, but I don't know how it compares to zbrush. Once you get comfortable importing and exporting back and forth from DAZ to a modeling/sculpting program to make your own morphs you'll want to do it all the time to fix wonky joint bends and clothing fits etc.
One thing that I've found is that having 3D scans of real people to use for reference for anatomical detail has really helped. For example, I can see that the way V4's arm bends looks wrong, but don't have the anatomical knowledge stored in my brain to see how to fix it. A 3D scan of a figure of similar body shape with her arm in the pose tells me what I need to know, and with the scan next to V4 for reference I make the fix to V4 and save it as an arm bend fix morph for later use.
Linvanchene, one thing to take into consideration is that rendering a 20,000,000 poly scene is not the same thing as posing a rigged 20,000,000 poly basemesh . It would be painfully slow to work with the figures at full resolution in DAZ. You can try it by turning on HD morphs and boosting the subdivision of a Genesis figure to subD 4 - not the final render subD which only comes into play at render time, but the live posing subD. You'll have to go into the top subD slider's preferences to remove the limit on subD levels available, I think it tops out at subD 2 by default. Be warned that raising it to 4 has a very good chance of freezing your computer and crashing DAZ. When rendering, DAZ can only handle that many polys because it's putting everything else on hold and using all system resources to do it with nothing available to compute bend morphs and collisions etc. That kind of shifting of resources isn't available while working on a live scene though, which is why there are two different subD modes available - a low one for posing and a higher one for rendering.
I think a 100,000 to 200,000 poly basemesh for Genesis 5 would be a winner though, a good compromise. The face in particular could use the extra detail and more facial polys wouldn't cause joint bending issues, plus it's already on a separate UV map so it doesn't need to have the same resolution as the body. If V4 has 70,000 polys and people were using her in 2006, I think it could be stepped up a notch for today's machines, rather than down (heck, V3 had even MORE polys than V4). It's useful having the Genesis figures available for background characters, but they look a bit rubbery and toony without HD morphs (which we users aren't allowed to make) or displacement maps (which have many practical usability limitations compared to morphs).
Does anyone know why DAZ decided to make the Genesis line with such a lower polycount than the earlier figures? Is it to simplify weight mapping and clothes autofitting? Or to make them usable in video games?
I do totally agree with you Linvanchene that a higher res basemesh is desirable - I'd be very happy with a 100,000 poly figure until our computers are powerful enough to play with even bigger ones.
I apologize... duplicate post
Low poly sculpting requires a change in workflow and tools. What is being discussed is way too much for general use figures. Took me a while but I have gotten the hang of making my base sculpts have the detail to work with subdividing without considering HD. There's plenty of tutorials on how to do this on the web. If you have zbrush, you definitely have what you need to make your characters without having a bloated 100K figure in a scene.
Also who would support two figures, a low poly that can be subdivided and applied a normal map and a super high poly figure that would lock the average user's workspace? There's no available resources for that.
While I like the face, IMO the arm resembles more of an orange peel with exaggerated bumps. It doesn't even look aged, as the skin isn't looser- it's just raised and some of it is not elongated into more wrinkled / loose pores but the raised part seems too much. The face and chest look quite respectable though!
Alrighty, let's note: V4 is showing 66,830 polys on my system. No subdivision at all.
So the idea that 100,000 polys would kill an average system is simply untrue. Just because the Genesis line are done off low-poly bases does not mean a higher poly base is unfeasable for general use.
The big reason for wanting a higher poly base is to allow for more detailed morphs than the low poly bases are capable of. As for support, I suppose the fact that V4 exists and to this day has new items made even if not sold here shows that the market can support a higher poly figure.
If we had HD morphs it wouldn't be an issue, but since we don't a higher resolution base would allow those who like to make their own morphs to do so.
Subdivision is a nice cheat to get around needing much mesh density, but it's just a cheat if you can't really morph the mesh like you would if the base had been that dense. Displacement and Normal maps are other cheats to get around the limits imposed by the first one.
If you're talking about subdividing a 100K figure even once, then it's an issue. Weightmapping a 100K figure is also an issue as it will have issues bending and setting up the map itself. It's easier to manipulate and weightmap lower poly figures. Then you add your details on top of that. That's where the subdividing and the normal map comes in. If you look at sculpting tutorials most start at working with the lower poly figures to your base shape then switching to the higher for details, then projecting that back to your lower poly base via either displacement or normal maps. That's really the change in workflow I mentioned. That's how all those zbrush figures started out; as lower poly meshes because lower poly is easier to work with to get your shapes. Considering those Zbrush artists have no HD tool, they use industry standard techniques to get their sculpts done, especially if they are using zbrush in conjunction with another tool such as 3DS Max or Maya. Those programs won't handle million poly sculpts like zbrush will, so that when they create their normal and displacement maps for use in those programs.
There are lots of tutorials out there as I said to do learn how to do this. If you can't get your morph out of 20k or less mesh, it highly doubtful you'll be able to get any more out of 100K. The number of polys really don't determine realism.
Yep. And as noted before, hobbyists were happily using V4 ten years ago, on machines that were even older... Add some clothing items and hair to V4 and we're already posing a 100,000 poly figure.
To me, that detail isn't bloat but quite useful. The lack of such detail in G3F is (I assume) one of the reasons the faces of characters sold for her look almost identical to one another. It's a cute girlish face, but a bit overly repetitive. Even among the G3F base figures, it's hard to see differences between the faces of Victoria7, Eva7, or Karen7 that can't be simply dialed into any of them. For sure, there are benefits to G3F such as much better bending than V4. I'm just hoping that the next line of figures could have the best of both worlds - the bending of G3F and the user accessible detail of V4.
G3M is also build off the same mesh as G3F (actually the meshes are different, but I understand the only real difference is the removal of a handful of polys from the breast area). Do you think G3M looks like V7 as well? I agree that a lot of the new G3F girls do look very similar to each other, but I do not think that is because the mesh is too low poly. Low poly might mean any morph would look less detailed, but you can still morph into radically different head shapes. There is a very interesting G3F character over on rendo called Kimama. It is of an older black woman, and looks nothing like V7 or G3F.
I confess I was also concerned about the lack of variety in the head morphs of the early G3F girls, in particular any based on V7 and some of the later new base characters. I am not sure why this is, but in the last few months a number of more distinctive G3F characters have arrived, both here at DAZ and elsewhere.
Frankly, very few people actually sculpted V4. They simply moved her face around or used dials. This is why when you subdivided genesis, it really wasn't that much of a difference between that and V4. Not a large number used an alpha, standard or pinch brush to do any type of detail. I think this is where most of the complaints about not having enough polys to make a sculpt because a lot relied on the move brush to adjust a chin or cheeks. Like I said, it's a different skill set to use low poly weightmapped figures. Large polys with weight mapping simply isn't efficient for posing or bending; that's why your normals or displacement is like the final pass in your render..
I think that really goes to the skill of the artist, not the amount of polys in a figure. You still need the tools and the change workflow to work with weightmapped figures. 100K is too much for a weightmapped figure to pose or animate efficiently, so you have to use techniques such as subdividing and normal maps to get the detail. zbrush also has tools to project details from a high poly to a low poly figure so you could still subdivide if you need more polys and project it back to the base genesis. I do this so my low poly and high poly keep in sync and also allows the figure to be used without higher details to save resouces as well. If I didn't have HD i could generate normals from that shape and not lose the detail. Since HD is out of reach for most people, learning how the industry does it is probably the answer to getting detail into your figures, along with using the proper tools such as mudbox, 3dcoat, zbrush, etc.
If I ever see a 20,000 poly DAZ figure that from a sculptural standpoint looks as unique and human as the best 70,000 poly V4 characters, I'll agree with you. I just haven't seen it yet. When looking at characters such as the lovely Dawn http://www.daz3d.com/dawn-for-v4-and-v5 that come with versions for both V4 and Genesis, and examining the promo images carefully, the V4 renders look a lot better to me than the V5 one. I remember being frustrated by that when those were new. The Genesis versions of Dawn and Summer Edition Liu on my computer just didn't look as finely sculpted as the promo renders - then I realized that most of the renders were of the V4 versions! Sneaky, sneaky was my thought, since I wasn't aware of the different generations of figures at the time...
I've got some experience baking normal and displacement maps from high poly sculpts onto low poly meshes. It's not an ideal method for things like muscle flex morphs on an athletic physique, unless I wanted to make a seperate displacement map for when the left bicep is flexed, when the right bicep is flexed,etc. That would be a bit overwhelming, and it was running into this very problem that brought me back to V4 and M4. With those figures, I have enough geometry to get the shapes I need right on the base mesh, no need for baking displacement maps for various limb bends and expressions which was my old technique for G2F and G3F.
Well right now displacement and normals are your answers to added detail, so that's really what people should be looking at. As talk about HD will end up geting the thread locked, as the status of HD is pretty clear, I'm going to bow out.
That's the lack of proportional scaling showing. V4 was pretty bad at that. The skin maps are just the wrong resolutions to look good with base maps.
The teeth in that image are a little too perfect, too...unless she has a dentist who specializes in cosmetic treatment, in her family, those teeth would cost a fortune.
I think part of the problem is what is the model going to be used for...
If it is a general purpose still/animated item then there have to be trade offs. And generally that is going to weigh more toward the posing/animating side of things. Which means the finer details will suffer/be lacking. But for static images, especially close ups, the lack of actual geometric details IS very noticeable.
One solution...set everything up, subdivide the mesh, export to modeller/zbrush and then scuplt the details...import the 'frozen' mesh back in and render. Unfortunately, that means no morphs and each render now has a unique, not rigged/resuable 'prop'.