3D Game Character - Bake or not to Bake?

IgnacJerema

Hello Polycounters!

      I am currently creating my first 3d Character, with "game ready" thought in my mind. I wanted to aproach it as a "pro" so I did my research about i.e workflow but still there are some things I did not quite understand. I want to make it as it should be done, that is why I am asking here. Please also correct me if I am wrong somewhere.

1.As you can see, I have currently just ended modeling stage and it is time for me to UV unwrapp character and then texture it, but as far as I know I should keep as low polygons as I could, with possible high quality if I'm thinking about games. Character is currently Low Poly with Subdivision Surfaces and Mirror Modifiers NOT applied.

And now I am wondering how THEY (pros) handle this moment. Should I be thinking about baking normals now? And if so, how should I aproach it?
Should I copy my character, then apply all modifiers on copy (making it ultra high poly) and bake normals on original low poly character? 

Assuming I am right, next question is coming to mind. How low poly should original charcater be?

Would you rather say:

• Low Poly
• Low, "but High enough to look good after baking" Poly
• High Poly, if game is going to be aiming in a realistic design

Thank YOU to everyone who is here to help! Cheers!

tahakitan

you really went far with the modeling. Most professionals including myself would do some low poly modeling than bring it into a sculpting program (usually Zbrush) and sculpt more details into it.  That will bring the character to the next level. after sculpting it, you would need to decimate it and bring it back into something like Maya and make sure that the low poly fits on top of it. If it doesn't than make the adjustment on the low poly.  than uv it and than bake it.  I would not subdivide it on a modeling program and call it good for the high poly, such as what you are doing.  Most character artists would see that and think it is not acceptable. for organic things and especially characters, you need to bake from a high poly sculpt, not a subdivision model. 

also, game characters don't use smoothing modifiers because that doesnt transfer to engine.  that is why we have to do baking.

pior

If you go the baking route, and assuming that you are using a realistic rendering style like what is done by default in all modern game engines, it means that the surface detail of the final asset will look just like that of your screenshot. Meaning that you will still have to put in a *lot* of work to get rid of the creepy plastic toy look that comes from not having any carefully crafted materials and textures applied. Check out the texture breakdown of this piece : 
https://www.artstation.com/artwork/mqO3de
https://cdna.artstation.com/p/assets/images/images/015/998/376/large/etienne-beschet-cha-vira-texturebreakdown.jpg?1550511009

But if you go the raw lowpoly route, then you can pull off a lot of oldschool tricks to take your asset into non-realistic rendering territory, going as far as overriding all realtime shading altogether.
https://80.lv/articles/001agt-004adk-sculpting-texturing-a-fierce-character-in-3d/

Put differently, if you go for a highpoly/bake workflow, then your target reference for the sculpt and paint job will be miniatures such as this :
https://i.pinimg.com/originals/a6/94/d4/a694d4815d5b3f5c83075716cf01c7bf.jpg

But if you go the lowpoly route, then your target could be as extreme as this :
https://i.pinimg.com/originals/11/3a/8f/113a8fccefef3e66fd9db00d8f2d1f38.jpg

There is no right or wrong way. However you do have to make a proactive choice at some point, because unlike a few years ago the potential level of detail is game assets is now virtually limitless , meaning that the risk to end up in "creepy plastic toy" territory is present more than ever as soon as you decide to go highpoly. Whereas a lowpoly only approach (along with the choice of using a non-realistic rendering style) can shield you from that by the very nature of it.

In both cases you do need to be familiar with what makes a character look cute in the first place, both in 3d and 2d. And handdrawn 2d studies are 100% the best way to understand that. If you can draw a cute character, then you'll likely be able to model it. Whereas if you tackle this study solely through modeling, the time it will take to understand these visual tricks will be much, much longer - because a model takes days/weeks to make, while a drawing takes mere minutes or at worse a couple hours.

Alex_J

You have to know what your goal is to answer these questions. 

A few things you got to keep in mind: 

>How much memory can the model use? That's mostly measured by how many vertices is it composed of, and how many textures does it require. In the end you have a total amount of memory that the model is going to use.

>What is the style I'm trying to achieve? This will determine what sort of textures you need, based on what shader you are using, in what engine, etc. 

So for instance if you said, "the model should be photorealistically rendered in unreal engine for a console game project," then yes, you probably do need to create a high poly sculpt that you would then bake a normal map from.

If you said, "the model should look like 2d and be used in a mobile game" then you can probably retopo what you already got to use much less geo, and spend more time focused on your textures and getting the perfect shape/silhouette. Maybe if your style or severe technical limits dictate you might not even use texture maps.

To say all that more simply, don't ask "what technique should I use?" Say, "this is exactly where I want to go, how do I get there?"

IgnacJerema

       Great thanks to You guys, tahakitan, pior, Alex_J. I saw your artstation profiles (except of Alex_J's, your artstation profile doesn't work ) and your characters are trully god tier.

       I'm gonna be honest as always. Topic is not crystal clear to me, yet you already made me learn a lot about character creation. Now I understand at least that even this post was approached by me incorrect way. 

This project is my first 3d character for my portfolio. I wanted to make it a similar way to someone who is working in game industry, to learn standards and workflow which would help me become a good artist and find a job a future.

I started creating this character following youtube speed-guide of https://www.artstation.com/thoredant artist. This is his tutorial character.


As always I wanted to follow it and be inspired by it, not copy it. Then I have found this art.

Learning from my previous mistakes, my goal was to create a character similar in style, to the one in the second screenshot (blue one). I assumed, I would end up somewhere between those two styles.

What I think, I know now about my character:

• the model should be stylized character
• the model should be rendered in blender for portfolio purpose

Now, my goal is to take this character to the end of process. What do you advise me, in which direction I should go with this project?

Any advices are pure gold to me, there is no one who could help me more. 

Alex_J

I wouldn't try to hit an unknown mark somewhere between two styles. Just choose one and copy it directly. You can start playing with your own style once you've got the basics locked down. 

I don't think either of those two character styles pictured above you need to bake normal maps. But other people would know  a lot more about those types of characters than me. If somebody can't tell you exactly what to do I'd spend a day researching similar style characters on artstation or sketchfab. Look for ones with 3d viewers and turn on the wireframe and different texture display modes. What is the topology like? What maps do they use? 
That will give you a better idea what sort of resources you'll need to finish your own character.