ORC- texturing. Day 01
by Radoslaw Marcin Kurczewski · 11/26/2006 (7:53 pm) · 15 comments
It was a royal pain, really. Making UV's by hand hurts. I've decided to do hands and legs repeatable ( there is no point in wasting more space then necessary) and use that space for head and torso. I've made first test render to check how well/badly lightmaps are getting baked out in new Maya. take a look at UV's and shadowmap (by no means final)

Test worked rather well so I've added plain colors for all main parts of the body. Most important trick later on will be separating parts belonging to various shaders (rubber, metal, cloth) because from what I see TSE can't really use one texture for all types of material. Sounds rather obvious but it took me a while to hammer that into my own head- it's one thing to know something from docs and something totally different to remember about it during work. Anyway- here is how low poly version looks with plain colors and shadowmap baked into them. Tomorrow some real texture painting...


Test worked rather well so I've added plain colors for all main parts of the body. Most important trick later on will be separating parts belonging to various shaders (rubber, metal, cloth) because from what I see TSE can't really use one texture for all types of material. Sounds rather obvious but it took me a while to hammer that into my own head- it's one thing to know something from docs and something totally different to remember about it during work. Anyway- here is how low poly version looks with plain colors and shadowmap baked into them. Tomorrow some real texture painting...

#2
One thing to keep in mind when modelling for TGE-A (new name for TSE):
You as an artist shouldn't concern yourself with the poly count when originally making the model. Since TGE-A is shader driven, use of normal maps and other shader capabilities allow you to model with millions of polys (if needed of course), and then reduce those polys by generating normal maps that will be projected onto a very low poly version of your model.
11/26/2006 (9:18 pm)
Looking good Marcin!One thing to keep in mind when modelling for TGE-A (new name for TSE):
You as an artist shouldn't concern yourself with the poly count when originally making the model. Since TGE-A is shader driven, use of normal maps and other shader capabilities allow you to model with millions of polys (if needed of course), and then reduce those polys by generating normal maps that will be projected onto a very low poly version of your model.
#3
That's almost bad advice. I don't mean to be rude but you fail to make clear to those who don't know about normal mapping:
1) The low-poly version needs to be made and rigged on its own.
2) The low-poly version is the only one which needs to be UV mapped.
3) It's considered easier by most artists (especially those who possess advanced tools like Zbrush) to make the low-poly version first and then add details for the high-poly version.
4a) The normal map can be painted by hand in programs like Photoshop or GIMP with normal-map generation plugins. (Draw the normal map in greyscale as a heightmap with light indicating highs and dark indicating lows.)
4b) Some artists think the easiest way to start with a texture is to subdivide your model, generate a normal map from it, and then bake it onto a texture.
5) In a low-poly model, it is only necessary to model what could be seen in a silhouette. Consider your design. If your modeling program allows you to, view your models without any lights. (Most programs let you do this.) Rotate the model around, and see what details to the shape of your model are lost. If they are lost when you do this, they aren't necessary for the low-poly model. Save them for the normal map.
(BTW: I wrote this for people who don't know much about making normal maps. I hope it's useful.)
11/26/2006 (9:59 pm)
@StephenThat's almost bad advice. I don't mean to be rude but you fail to make clear to those who don't know about normal mapping:
1) The low-poly version needs to be made and rigged on its own.
2) The low-poly version is the only one which needs to be UV mapped.
3) It's considered easier by most artists (especially those who possess advanced tools like Zbrush) to make the low-poly version first and then add details for the high-poly version.
4a) The normal map can be painted by hand in programs like Photoshop or GIMP with normal-map generation plugins. (Draw the normal map in greyscale as a heightmap with light indicating highs and dark indicating lows.)
4b) Some artists think the easiest way to start with a texture is to subdivide your model, generate a normal map from it, and then bake it onto a texture.
5) In a low-poly model, it is only necessary to model what could be seen in a silhouette. Consider your design. If your modeling program allows you to, view your models without any lights. (Most programs let you do this.) Rotate the model around, and see what details to the shape of your model are lost. If they are lost when you do this, they aren't necessary for the low-poly model. Save them for the normal map.
(BTW: I wrote this for people who don't know much about making normal maps. I hope it's useful.)
#4
11/26/2006 (11:47 pm)
It is helpful. Thx!
#5
You should try 3D max pelt mapping feature. Since I'm using that, my UVW's have improved (no stretching), and... I make them 10, maybe even 20 times faster now.
[edit]... just see you use Maya, s nevermind[/edit]
11/27/2006 (12:35 am)
Quote:Making UV's by hand hurts
You should try 3D max pelt mapping feature. Since I'm using that, my UVW's have improved (no stretching), and... I make them 10, maybe even 20 times faster now.
[edit]... just see you use Maya, s nevermind[/edit]
#7
No problem at all--thanks for the alternate viewpoint! Not being an artist, I can only go with what I'm told ;)
11/27/2006 (5:54 am)
@Brian:No problem at all--thanks for the alternate viewpoint! Not being an artist, I can only go with what I'm told ;)
#8
Back to polycount/normals thing- I do not use simple subdivision method and not because it's bad but rather because it's not sufficient (to me that is- can't say a thing about other people). In my case it will be always creation of low poly model, then- based on this LP geometry (to make sure I have all envelopes right) I am building high poly model with all details that would be bigger then 5 mm in real life on geometry (no textures on this stage). From this- in most of the cases deadly heavy (from 100 K to... well... whatever is needed)- model I am baking normal map to be applied on low poly one. I created UVs now as making them on high res model would be a torture and I am not really into self mutilation, thank you very much. Having them in place now means that further normal map will fit exactly... or so I hope... Anyway- for now I will concentrate on adding bit more details to diffuse texture (got to do something for pleasure for exchange).
11/27/2006 (6:36 am)
Thanks a lot guys- I never expected to get so much attention. Back to polycount/normals thing- I do not use simple subdivision method and not because it's bad but rather because it's not sufficient (to me that is- can't say a thing about other people). In my case it will be always creation of low poly model, then- based on this LP geometry (to make sure I have all envelopes right) I am building high poly model with all details that would be bigger then 5 mm in real life on geometry (no textures on this stage). From this- in most of the cases deadly heavy (from 100 K to... well... whatever is needed)- model I am baking normal map to be applied on low poly one. I created UVs now as making them on high res model would be a torture and I am not really into self mutilation, thank you very much. Having them in place now means that further normal map will fit exactly... or so I hope... Anyway- for now I will concentrate on adding bit more details to diffuse texture (got to do something for pleasure for exchange).
#9
Also keep in mind, your using quads, each quad will be triplet into a triangle if you intend to use it in the engine. So, take your total poly count and x2 it.
11/27/2006 (1:24 pm)
Unless the shape, or textures are going to be unique, left to right. You should remove half of the model, UV it (I didn't say texture yet) then mirror the two halves, then texture it.Also keep in mind, your using quads, each quad will be triplet into a triangle if you intend to use it in the engine. So, take your total poly count and x2 it.
#10
11/27/2006 (2:23 pm)
As mentioned before- legs and arms mirrored, rest unique (I do not really like this "mirror like" characters). Polycount is given in no. of triangles (in gaming does not make sense to give polycount for quads, really) so it will not change- I am showing quads as it allows to judge the mesh.
#12
11/27/2006 (4:32 pm)
Looking good! Good experience to go through, I ended up having to crunch something like this out last august, was interesting.
#13
To the subject- I had to do that kind of stuff for customer working with HL2 engine about a year ago and- to my surprise- I've discovered that TSE has much better normal map effects. In HL2 they were almost like a bump- here they are what I call proper normal map.
NO, it is not another X vs. Y war- just opinion on one particular function
11/27/2006 (5:14 pm)
Proud to see you here, Timothy- you are THE MAN when it comes to Torque graphic. To the subject- I had to do that kind of stuff for customer working with HL2 engine about a year ago and- to my surprise- I've discovered that TSE has much better normal map effects. In HL2 they were almost like a bump- here they are what I call proper normal map.
NO, it is not another X vs. Y war- just opinion on one particular function
#14
01/08/2007 (12:03 pm)
How do I get the polyCount? The console command 'metrics(Interior);' doesn't give any valid information.
#15
01/08/2007 (12:43 pm)
Ekhm... polycount of... ? 
Torque Owner Edward Smith
Silencersoft
I think its best in TGEA to use different textures for different parts.
Look forward to seeing this in TGEA :-).