Map baking research and things i could not figure out
polycounter lvl 4
Hi everyone, i am new to this forum even though i have been in and out of topic in here through google search for quite a time, i dont know why it took me so long to register after all (i assume desperation); well anyways i'm here; so hello 
I want to cut short and be on point as much as i can.
My problem is that: i have been doing cgi related stuff for 3 years, and i am very dedicated about it and has been the focus point of my life ever since, i think i had the grasp of most of the things i had studied and trained my self about it through the things i learned online, since i did my major about something completely unrelated. So; at this point of my life, i decided to up my projects a little and really focus on creating models that i can show case in my portfolio, i started modelling a car and i have finished the most of it. Here is a snapshot:
This is a test render i took with Arnold in Maya. Now i want to texture this model in Substance Painter and optimize it every aspects. But even though i had watched many tutorials about map baking principles, theory and workflows in various resources. I have watched Allegorithmic's tutorials, and other extensive tutorials about map baking and i already had some theoretical information about the subject. However, my results are very off from what is expected. But i guess i am having trouble about some principle things about map baking and i guess there are some personal things that i overlook and could not get the grasp on. I duplicate the workflows of some people i watch online with very simplistic models i created but even then i fail. Believe me i feel stuck, untalented and foolish. So decided to note down what i realize to raise up in my mind through my process.
- I feel confused about what a normal map can achieve and what it can not. I know how maps are created through a envelope and we may assume like there is ray trace sent from this envelope to the mesh, and a map is created based on the normal of a surface, kinda like bump maps and displacement maps.
To exemplify my confusion here: When i created the model ive uploaded, i modelled the mesh considering that the end result will be the preview model. But what it actually does is to preview a smooth model with more polycount, it is not the end result. Soooo, can i created a smoothed version of the model and created a normal map to achieve the previewed look with the un-smoothed, low poly mesh. I mean can normal maps achieve corner smoothness i have created with bevelled edges. I assume not. I am not sure. I have seen many people baking normal maps but i don't know how their decision making mechanisms worked when they were modelling. I guess i am having this problem because i am having trouble with the mechanics of map baking. Like i had another confusion when i see how substance painter imports maps from high poly to low poly. You think you do not preview an envelope but you enter the enveloped distance you want to create.
I am also not sure how curvature maps are created, is it throguh the UV seams or does it have another mechanism that detects hard edges.
I may have written a lot, but i am writing here my stream of thoughts just because you may be can see the problem in my mindset.
Thank you in advance whoever took his time to read this. You dont have to write a specific answer to my problems, any research material or tutorial that you think that will correct me will be much appreciated. I am also willing to investigate very technical, boring written online material. Event some key terms and critical concepts that you may suggest me will be helpful.
I want to cut short and be on point as much as i can.
My problem is that: i have been doing cgi related stuff for 3 years, and i am very dedicated about it and has been the focus point of my life ever since, i think i had the grasp of most of the things i had studied and trained my self about it through the things i learned online, since i did my major about something completely unrelated. So; at this point of my life, i decided to up my projects a little and really focus on creating models that i can show case in my portfolio, i started modelling a car and i have finished the most of it. Here is a snapshot:
This is a test render i took with Arnold in Maya. Now i want to texture this model in Substance Painter and optimize it every aspects. But even though i had watched many tutorials about map baking principles, theory and workflows in various resources. I have watched Allegorithmic's tutorials, and other extensive tutorials about map baking and i already had some theoretical information about the subject. However, my results are very off from what is expected. But i guess i am having trouble about some principle things about map baking and i guess there are some personal things that i overlook and could not get the grasp on. I duplicate the workflows of some people i watch online with very simplistic models i created but even then i fail. Believe me i feel stuck, untalented and foolish. So decided to note down what i realize to raise up in my mind through my process.
- I feel confused about what a normal map can achieve and what it can not. I know how maps are created through a envelope and we may assume like there is ray trace sent from this envelope to the mesh, and a map is created based on the normal of a surface, kinda like bump maps and displacement maps.
To exemplify my confusion here: When i created the model ive uploaded, i modelled the mesh considering that the end result will be the preview model. But what it actually does is to preview a smooth model with more polycount, it is not the end result. Soooo, can i created a smoothed version of the model and created a normal map to achieve the previewed look with the un-smoothed, low poly mesh. I mean can normal maps achieve corner smoothness i have created with bevelled edges. I assume not. I am not sure. I have seen many people baking normal maps but i don't know how their decision making mechanisms worked when they were modelling. I guess i am having this problem because i am having trouble with the mechanics of map baking. Like i had another confusion when i see how substance painter imports maps from high poly to low poly. You think you do not preview an envelope but you enter the enveloped distance you want to create.
I am also not sure how curvature maps are created, is it throguh the UV seams or does it have another mechanism that detects hard edges.
I may have written a lot, but i am writing here my stream of thoughts just because you may be can see the problem in my mindset.
Thank you in advance whoever took his time to read this. You dont have to write a specific answer to my problems, any research material or tutorial that you think that will correct me will be much appreciated. I am also willing to investigate very technical, boring written online material. Event some key terms and critical concepts that you may suggest me will be helpful.
Replies
The sticky threads about normal maps on this very forum will tell you almost everything you need to know about getting a good bake.
Also,
Don't use that car as a test case, its complex and will not be 100% straightforward even with experience.
If you feel foolish now you just wait until you learn it, i felt the same way when i was trying to learn about baking but looks so simple after you understand the basics.
Learn about tangent space and hard edges, those were my biggest problems and questions about it pop in this forum all the time.
Try out marmoset toolbag, fast interactions and easy to set up, best baking tool in town.
These days, trouble shooting a bad bake is not a big thing at all anymore, though it used to have me pulling my hair out. Also, Marmoset has fantastic tutorials on their website. Even if you don't use Toolbag, they are very visually oriented tutorials that may help if you haven't looked at them.
https://marmoset.co/posts/toolbag-baking-tutorial/
About what normal maps do, they only change the way light interacts with the normals of an object. So, surface level details only -- no silhouette changes can be achieved with normal maps. It's like putting on make up, in a way. If you're a bit doughy, the make-up can't hide the double chin (nor make a double chin exist where there isn't one.) You just got to play around with different models to learn when normal maps are useful, and when you ought to use geometry for detail.
Curvature maps look at the angle of the surface. You can usually adjust to what degree, and you can of course adjust the histogram levels in photoshop or allegorithmic programs. This is useful for targeting metal edges to receive wear, or leather, etc.