Official 3D art thread - Warband

Users who are viewing this thread

Well, here's a typical test-case.  Burned AO from Barf's skeleton (so we can't blame my skills as a modeler, lol):

Before AO:

skele_ao01.jpg


After AO.  Looks all right at a distance, but up close, there are issues.

Red areas are where we have anomalous results.
skele_ao02.jpg
skele_ao03.jpg
skele_ao04.jpg
skele_ao05.jpg
 
oh wait, you're baking in xnormal as if it has a cage? that looks like misprojections. You shouldn't need to set up a cage and projection to bake ao straight from an LP model to an LP model. Any respectable modeling package can do it without that.

If you're not projecting it, then i would recommend scaling the mesh up ~100x or so, and see if the result is different.


edit: and it would be good to show the AO without diffuse lighting. most the visual ****tiness on that mesh are just poor shading around extreme edges with averaged normals.
 
oh wait, you're baking in xnormal as if it has a cage? that looks like misprojections. You shouldn't need to set up a cage and projection to bake ao straight from an LP model to an LP model. Any respectable modeling package can do it without that.
Can I get a little more explanation about that?  I don't use Max, so the concept of "cage" is a bit fuzzy to me :smile:

Anyhow, I just loaded up LP --> LP --> bake, after resizing up 10X.  I'll fire it up without any lighting, but that stuff I'm pointing out is in the burn. 

It's a good example mesh for this stuff because it is trying to do some fairly extreme things with the LP geometry and it's a complex AO subject.

I'll try 100X, see how much difference it creates.  Also need to figure out how to keep xNormal from maxing out my quad-core, it overheated my poor CPU when I started a burn with 1024 rays  :lol:
 
You shouldn't need that many rays.

You kind of have the wrong idea about xN. It's for projecting bakes - using a lowpoly to bake detail from a different mesh. If you didn't set a custom cage, the xN defaults for "ray distance" or something like that, will be used. This is to replace a cage and ensures rays are cast from outside the source mesh you wish to bake onto the LP. Otherwhise you have ray misses.

What it looks like here is there is some default amount by which the rays origins are pushed out to clear what would usually be a highpoly mesh with a completely different structure... So the rays are passing through the other parts on their way and baking them down onto whatever faces they end up hitting.

If you decrease the value for ray distance (i forget what it's called or where it is in xN) you will probably get better results - but its better to just do a standard render to texture than a projected bake in this case, anyway... Which afaik xN can't do (not that it should - it's not designed for that purpose.)
 
Resizing 100X helped quite a bit; there are still a few minor errors but nothing too bad. 

Strange that it doesn't automatically scale it up for you if radius < 2.0.  I'll make a feature request or three :smile:

It really can't deal with anything re-using uv space, though.  I just did a burn with some armor and the results were pretty awful.

That's a technique I use a lot for hard-surface stuff where I need to conserve space, amongst other things.  Are there any workarounds?
 
xenoargh said:
That's a technique I use a lot for hard-surface stuff where I need to conserve space, amongst other things.  Are there any workarounds?

offset the stacked portions by 1
 
I'm not sure what you mean there.  You mean move the uvs some minimal distance?  Or move all of the stacked stuff off the chart 1.0 u or v?

For example, if I have a bunch of spikes sticking out of an armor... they're all the same area of the chart, but I may have displaced their geometry a bit, they're at different angles, etc., and some of the geometry intersects the armor by varying amounts. 

My results with that aren't good; I'm expecting it to average it across each copy and arrive at an average result, which won't be perfect but it'll be OK.  Instead, I get a pretty extreme result.
 
xenoargh said:
move all of the stacked stuff off the chart 1.0 u or v

xenoargh said:
My results with that aren't good; I'm expecting it to average it across each copy and arrive at an average result, which won't be perfect but it'll be OK.  Instead, I get a pretty extreme result.

Well, that's a limitation of how you've uved it. Even with hand painting the best you could hope for is imperfect.
 
Harry_ said:
Well, that's a limitation of how you've uved it. Even with hand painting the best you could hope for is imperfect.
This: sounds like the algorithm's design in that particular situation is best-effort, which is quite the misnomer for absolutely nothing.
 
Woah, yoroi starting to reach the '50% done' mark.

Started doing the knots, strings and lacings.

Way to go, but I'm very happy with the results so far:

6996259114_1ae74db537_c.jpg
6996258846_96516e8ac0_c.jpg
7142347489_49533fbb46_c.jpg

 
hey guys this is what I've been working on on past 3 weeks:
Render_WIP_3.jpg

OpenBRF_Pre-view.png
it's really wip and it's about 20k polys.I can't whait to start texturing and baking it.

and here are 2 other ships that  I did for a warband project witch will probably never go somewere:
render10.jpg
 
atolest said:
hey guys this is what I've been working on on past 3 weeks:
Render_WIP_3.jpg

OpenBRF_Pre-view.png
it's really wip and it's about 20k polys.I can't whait to start texturing and baking it.

and here are 2 other ships that  I did for a warband project witch will probably never go somewere:
render10.jpg

Absolutely amazing! You should try modeling the Vasa once, in particular its rear should provide you with a nice challenge :smile:
 
atolest said:
hey guys this is what I've been working on on past 3 weeks:
Render_WIP_3.jpg

OpenBRF_Pre-view.png
it's really wip and it's about 20k polys.I can't whait to start texturing and baking it.

and here are 2 other ships that  I did for a warband project witch will probably never go somewere:
render10.jpg
Seems like you moddeled to much of it, You could probably bring it down to 20k with good alphas.
 
OrangeKnight said:
atolest said:
hey guys this is what I've been working on on past 3 weeks:
Render_WIP_3.jpg

OpenBRF_Pre-view.png
it's really wip and it's about 20k polys.I can't whait to start texturing and baking it.

and here are 2 other ships that  I did for a warband project witch will probably never go somewere:
render10.jpg
Seems like you moddeled to much of it, You could probably bring it down to 20k with good alphas.
you are probably talking about those ropes that goes to the mast,well Im not planning to bake it into alpha maps because I'm planning to make that ship a boarding battle,and I want it to be quite detailed.
I will keep doing that ship with no really worry with polygons(or at least detail),because I am also planning implementing it in unreal engine or cryengine as other of my personal project
 
atolest said:
OrangeKnight said:
atolest said:
hey guys this is what I've been working on on past 3 weeks:
Render_WIP_3.jpg

OpenBRF_Pre-view.png
it's really wip and it's about 20k polys.I can't whait to start texturing and baking it.

and here are 2 other ships that  I did for a warband project witch will probably never go somewere:
render10.jpg
Seems like you moddeled to much of it, You could probably bring it down to 20k with good alphas.
you are probably talking about those ropes that goes to the mast,well Im not planning to bake it into alpha maps because I'm planning to make that ship a boarding battle,and I want it to be quite detailed.
I will keep doing that ship with no really worry with polygons(or at least detail),because I am also planning implementing it in unreal engine or cryengine as other of my personal project
Just because something is going into one of those engines doesnt mean you should waste polygons. Its a bad practice and some of those details nobody would notice the diffrence. Look at docm's ships, they look great and are very efficient.
 
Orange... it's not a waste of polygons. He decided to put details into it, and he did a great job. Alpha maps for something that big require GIANT texture sizes just to make it work properly.
 
...and a huge number of alpha tests in a lot of views, unless you use almost as many triangles.  Alpha tests get fairly expensive when you're talking about most of the screen, perhaps multiple times.  And there are other issues with that.  Basically, alpha trans is great, but mainly for small stuff that won't cover a lot of the screen and where saving triangles is more important than per-texel evaluation.

A 20K model for something like that is very reasonable, basically, if anything, he might want to spend more on areas like the deck so that he can use vertex AO, because it'd really help performance if as much as possible can be tiled or use mirroring.
 
Back
Top Bottom