"Yes, but how many polygons?" An artist blog entry with interesting numbers

Another thing to consider with Uncharted, and a lot of other game assets, is that realtime graphics have certain limitations that dictate the technical aspects of the models.

GPU skinning is very cheap and memory efficient, whereas blendshapes are slower and eat up RAM very fast. Our facial rigs using 30-50K quads on the face, about 15K on the teeth and a few more thousand on the eyes and such come out at about 300MB, and that's just the geometry and rigging data without textures and such.

Most game facial rigs are using bones and maybe a few blendshapes on top of it to fix complicated cases. Now painting the skin weights for hundreds of bones is incredibly time consuming, but you can save the weight values for a model and reload them. So, you can take the same base mesh and modify it to fit a different face and then load the weighting and save a lot of time. You'll also get to keep the UVs which allows you to re-use texture data. Games with crowds can use this to share texture maps across an entire range of characters; for example the roughness (the size of the specular highlights) is usually very similar on every face, as it depends on general anatomy like the thickness of the soft tissue and the usual amount of oil on the skin.

Another thing we usually do is that the facial wrinkles are all modeled into the face, things like the folds on your brow, the 'laugh lines' going from your nostrils to the corners of the mouth, or the crow's feet around the eyes. This allows us to sculpt all these wrinkling and folding into the blend shapes, so you'll actually see the skin deform; and this is also visible in the Maya viewport of the animators. But these wrinkle patterns are very much like fingerprints, while there are some common elements it's still specific to the individual. I also think that they do change with age - I've been staring at my face in the mirror making grimaces for the better part of a decade and I'm looking different nowadays. But I'd say that after about 35-45 years, it gets fixed and starts to wear in with old age.
Games usually don't have this, but they're using blended normal maps instead and those are independent of the underlying polygon structure. So they have no problem reusing the same model.
 
The question remains... Yes, but how many polygons??

Indeed that is the question, if there was any time to talk about it, it would have been in the "modeling of Nathan" conference. my guess was they probably felt it wasn't necessary, and what they were doing with the character was more important. (animations, effects, ect) which is fine, they opened up to the public more than many expected in this point in time, kudos for at least doing this.

------------------------------------------------------------------------------------------------------------

I was hoping to see more Qauntum Break Details by this time. it's soon to be 2015 and we haven't seen many other characters.
 
Claire, cutscene model, Resident Evil Code Veronica.

I13wk1U.jpg
 
I've just started messing around with 3D modeling using Blender. It's give me a whole new appreciation for how great games like U4 look considering it takes my 750Ti around five mins to render a basic character and they are rendering Drake, all his enemies and an incredible environment 30 times every second in real time !

Laa-Yosh, I take it from previous posters comments that you work in the movie industry ?, any tips for someone just starting out modeling who wants to do it as a career ?. What programs would you suggest I learn, I take it you don't use blender ;).

Cheers.
 
I only work in CG animation, at Digic Pictures - we're usually making trailers and cinematics for games, like AC, the new COD or Mass Effect. But I do have a lot of pals around the world in VFX studios too :)

As for working as a modeler, the most important aspect is not the software - that can be relatively easily learned, and it's constantly changing anyway. Nevertheless, most studios are using Maya, or Modo, and of course Zbrush or Mudbox for sculpting. But the general principles are the same, so learning Blender would help as much as anything else.
The more important aspect is traditional art skills, may it be anatomy or industrial design or whatever; and having a good eye. It's a bit complicated to define but it has to do with sight being a skill instead of just a sense, the ability to understand shapes and forms and silhouettes, balance, rhythm, basically the things that make stuff look 'good' or in other words, aesthetically pleasing :) This is the harder thing to learn, takes years of drawing and such and there really is no good replacement for this.
 
CGTalk.com should be able to provide good resources for someone just starting out with 3d modeling. Sorry to plug in a different forum, but the people there have a wealth of experience.
 
I only work in CG animation, at Digic Pictures - we're usually making trailers and cinematics for games, like AC, the new COD or Mass Effect. But I do have a lot of pals around the world in VFX studios too :)

As for working as a modeler, the most important aspect is not the software - that can be relatively easily learned, and it's constantly changing anyway. Nevertheless, most studios are using Maya, or Modo, and of course Zbrush or Mudbox for sculpting. But the general principles are the same, so learning Blender would help as much as anything else.
The more important aspect is traditional art skills, may it be anatomy or industrial design or whatever; and having a good eye. It's a bit complicated to define but it has to do with sight being a skill instead of just a sense, the ability to understand shapes and forms and silhouettes, balance, rhythm, basically the things that make stuff look 'good' or in other words, aesthetically pleasing :) This is the harder thing to learn, takes years of drawing and such and there really is no good replacement for this.


Thank you very much for your reply mate, very interesting, I will try and improve !
 
From Famitsu interview with FF XV devs:
There are about five million polygons per frame, witch each character made up of at max 100,000 polygons. The inner hair alone has about 20,000 polygons, which is five times the previous generation.

Character models have around 600 bones, which is roughly 10-12 times greater than what was seen last generation.
 
Back
Top