Cinematic models

Lephenix

Hello fellow polycounters, i wanted to ask you some questions, because i'm really interested in cinematics and movies more than video games, although i love video games and modding.
The questions are: What's the difference between a game model and a cinematic model? What are the steps to make a cinematic model?

I know for making game model is generaly:

1) Base Mesh/ Z-Spheres
2) High Poly/ Sculpt
3) Retopo
4) Low Poly
5) Unwrap Low Poly
6) Bake details of high poly on low
7) Texture
8) Render/Post Process

For a cinematic character is it:

1) Base Mesh/ Z-Spheres
2) High Poly/ Sculpt
3) Retopo
4) Unwrap
5) Polypaint
6) Texture
7) Render/ Post Process

I ask that for anything like environments, characters, vehicles, weapons... Thanks in advance.

I already got one answer:

7mch wrote:

Well,I have no idea about environments..

As for characters, I guess the workflow of modelling and texturing is very similar to that you described.

When it comes to a pre-rendered cinematic the major difference is that you can have

• Dynamic hair system
• Cloth simulation
• Subsurface scattering shader for skin, teeth and eyes (Refractive shader for the cornea of the eye)

And off course you can render your imagery in multiple passes.

All available technologies such as cloth and hair simulation, muscle system, advance skin shaders, global illumination and image based lighting solutions can be used to create the final picture. Any element of the character such as hair, eyes or clothing can be rendered in multiple passes allowing a greater control during post production.

No limits on poly-count, as long as the polygons are used sensibly. However any cinematic character should be animation friendly. That means that the topology of the character is very important to provide good deformation for realistic animation.

There are also no especially restrictive limits when it comes to texture maps and texture sizes.
Andrew Whitehurst from Double Negative who came to our Uni to do a presentation about the techniques used in "Scott Pilgrim vs. the World" film mentioned that they would usually have about 10 to 15 UV layouts for each character. There will be a separate map for face, eyes, hands, clothing etc. Moreover, each UV set would have colour, specular, displacement and more other maps where each of them would be somewhere between 4k - 8k.

I just want to regroup many infos as possible, so if you have other advices, techniques, limitations ... I am open , would be cool if some cinematic artists who work at Blur Studios or something like that could answer, that would be great. Thanks in advance .

ZacD

I don't know if this is standard practice, but from some breakdowns I've seen, they create subdiv models with displacement maps, so it will animate well, but they can get close ups without seeing harsh lines. I've seen some stuff where they set up the characters to "lod" by have the character subdivide as the get closer to the character, and have the displacement kick in later, so save a bit on render time and view port performance.

Lephenix

Ok, thanks for the answer.

gray

your asking a really broad question that is very difficult to sum up in a few lines. also every studio is going to have its own slight differences. and each production has a different look that requires different approaches and tools. so there is no definitive answer.

the quote from 7mch is about right.

if you want to do a character for a cinematics/film the first things you should do is learn how to use displacement maps, multi-uv tiles and get comfortable rendering in mental ray. try to find some examples of resurfacing for film, screenshots etc. that should take a good bit of work.

the rest of the pipeline for cinematics/film is really hard to explain. so i would not worry about it. you might want to look into character deformation setup. making facial blend shapes for animation. hair, hair dynamics, sub surface scattering, etc, etc, etc. there is huge list it will take a long time.

Kwramm

you might want to search for threads from vargatom, he works for digic and really knows what he's doing. Lots of interesting insights there.

The process of just modeling is fairly similar - the differences are in the details themselves and that there's just more complexity to everything - shaders, rigs, textures, etc. E.g. you have special skin shaders or shaders for other matierals. Those take additional types of maps than just your standard normal/diffuse/spec. Or that you're not baking in AO in general, etc, etc.

Some studios have different versions of the same character, depending on the shots. In film you can often plan your shots ahead and then model and rig the character (or multiple versions) for the best possible result. (i.e. for a hand close-up just model 1 hi-res hand assuming you never get close to the character for the rest of the clip). In games characters have to be pretty flexible because the same model has to support all the actions the player may possibly want to do - or the player can zoom as close or far away from the model as he wants.

I think Gnomon has some interesting DVDs about modeling/rendering for pre-rendered where they explain the subtle differences.

Lephenix

Thanks for allt the infos, it's way more difficult than i thought, i'm still a "beginner", so i'll have the time to practice all those.