Understanding the pipeline
polycounter lvl 4
Hello guys!
I'm pretty much still a beginner and I was always curious about the pipeline for making cinematic/movie ready models.
As far as my understanding goes the pipeline for games goes like this (lets just take a model of a human for instance)
But I was always wondering how does it all work for something other than PBR rendering. I'm learning 3D for almost 2 years now and I've never rendered anything in a ray-based renderer (it that what you call them? I mean: mental ray, Vray, arnold, Corona, you name it).
I've been working on models for 3D printing and some mobile games, so I've never came across it. So the whole process is a huge mystery to me.
I know absolutely nothing about shaders, the maps you need for them, where and how do you create them and stuff.
So to test out the waters I've decided to try and make a model for all of those purposes. It may not be a cinematic quality model, I'm pretty sure it's not, but it is purely for research purposes. So I ended up just making a some kind of superhero body:
And the first cornerstone I've stumbled upon is retopology. For some reason I can't google enough information due to my lack of knowledge of the subject. How do you retopologize a model ment to be in, for example, a movie? I mean, you cant get away with a model that consists of 10k polys like you would do for a game for instance. But retopoing to a higher amount of polys seems impossible to me. What I came up with was to make a low-poly with a good topology, import it in zbrush, divide it a couple of times, project all the details and take the 2nd subdiv. I don't know if it's legit or not. It would be awesome if someone could explain me more and share some information about this.
Thank you!
I'm pretty much still a beginner and I was always curious about the pipeline for making cinematic/movie ready models.
As far as my understanding goes the pipeline for games goes like this (lets just take a model of a human for instance)
- Making highpoly sculpt (with all the details and pores etc)
- Retopologizing (you end up with not really dense model)
- Baking highpoly details onto retopoed mesh
- You end up testuring it in Substance Painter or DDO or something like this
But I was always wondering how does it all work for something other than PBR rendering. I'm learning 3D for almost 2 years now and I've never rendered anything in a ray-based renderer (it that what you call them? I mean: mental ray, Vray, arnold, Corona, you name it).
I've been working on models for 3D printing and some mobile games, so I've never came across it. So the whole process is a huge mystery to me.
I know absolutely nothing about shaders, the maps you need for them, where and how do you create them and stuff.
So to test out the waters I've decided to try and make a model for all of those purposes. It may not be a cinematic quality model, I'm pretty sure it's not, but it is purely for research purposes. So I ended up just making a some kind of superhero body:
And the first cornerstone I've stumbled upon is retopology. For some reason I can't google enough information due to my lack of knowledge of the subject. How do you retopologize a model ment to be in, for example, a movie? I mean, you cant get away with a model that consists of 10k polys like you would do for a game for instance. But retopoing to a higher amount of polys seems impossible to me. What I came up with was to make a low-poly with a good topology, import it in zbrush, divide it a couple of times, project all the details and take the 2nd subdiv. I don't know if it's legit or not. It would be awesome if someone could explain me more and share some information about this.
Thank you!
Replies
1st of all, there are not many differences between a real-time engine and a offline renderer except the latter gives more accurate results because everything is pre-calculated. Both sides can be physically based. They sometimes use different maps to achieve the same goal. For example, v-ray use a specular/glossiness map while ue4 use a roughness map. Same thing, just 2 different ways of doing it.
For characters, unless i'm mistaken, I don't think they use low-poly with normal maps on their characters. At least not the AAA studios. That seems too low quality imo. They probably brute force super high poly models textured with ptex for example. Everything is pre-rendered so it's not an issue!
The normal map/ retopo/ low poly is more of a gaming industry workflow because of the constraints of running in real-time...
Concerning retopo, i don't think it's harder to retopo a 1 million poly object than it is to retopo a 50k object. You just draw new polygons on top of your high res object. In zbrush you could try zremesher for retopo. Or learn something like topo-gun.
Oh, I guess I wasn't clear enough what I meant. I was talking about the final mesh, not the object that you retopologize. I mean, that is not new for me to retopo. But there is a difference in making retopologized mesh like this
And like this
It is really hard for me to believe that is was all done by hand, so I tend to think that mine approach (which I'm pretty sure I can't call "mine", it is just something that I came up with on my own, don't really know if it's the right thing to do or not) with low-res retopo and projection is more or less legit.
I don't know if i'm answering your question haha...
In your portfolio work, you can do what you need to do to get that nice bake or the look you want. However ignoring the practical realities of development is going to set back your work. At some point you will need to understand how to create efficient low poly models (and why).
this question is confusing because you are asking about film models and showing shots from film models. Now it is true that game models are getting closer to film models every day, and that the amount of polys you can put in a character is crazy - but the way game models are getting close to film quality is down to being clever with the model, the additional map and neat shader tricks.
What I want to say here is really this: If you see a shortcut, try and understand what the shortcut is skipping over and why.
Making your low poly model is skill, traditional modelling is a skill. It's not too scary and like zbrush and anything else, it gets faster and easier with practice and understanding. You can get really far with automatic topology options but there's alot more to game art than sculpting and baking.
Keep reading, keep making things, keep looking for help when you run into problems and you will pick up this stuff! There is no one golden rule for what is and isn't a good final game model. (other than looking good and running smoothly!)
But anyways thanks for your reply!
I don't think a ''non-pbr renderer'' is a thing really. I'm pretty sure all big renderers are physically correct. You can trick a renderer to become non-physically correct though! There are some things that simply can't be calculated in real-time. Complex refractions/caustics for example. This is something that has to be rendered in a movie. For a movie they just go all-in. High poly models, huge uncompressed textures, complex materials/shaders, physic simulations, etc.
http://polycount.com/discussion/69005/character-polycounts-in-film
Actually it's exactly the same, we still use a lower resolution model and add a displacement map on top of it.
You can't work with a 20 millions poly mesh.
During the render the engine subdivide the mesh many time to give enough geo to have accurate displacement, but we still have a " low " poly.
Again you'd be surprised, you should not forget that in a film there is a hundred people working on a shot, they have plenty of time to do retopo of everything. Same for UV, I won't be surprise if I find out that they indeed unfold every object in a transformer battle scene.
Cheers !
I've never used toolbag, is it only a real-time renderer?
Thanks for the clarifications Burpee!
I might be a little bit late with this one, but I found an example
Here's a model I've made a while ago. If I were to make this helmet a, per say, game model, the mesh would look just like a general helmet shape, all of the parts probably in one piece. All the details would be on the baked normal map. But if I wanted it to look good in closeup, if I wanter to render it out in V-ray, do i still retopo just the main shapes? Should you break it all up in pieces? I am really sorry if my questions are really dumb, I wish I could find some timelapse or making of with the workflow around this.
gl !