Home Technical Talk

Baking Issue

polycounter lvl 13
Offline / Send Message
Wesley polycounter lvl 13
Feel a bit of noob here, and also kind of embarrassed to be contributing to the huge amount of baking threads that sit on Polycount.

I spent a good day or so reading up as much as I could about normal mapping and procedures that are used. And tonight I tried baking down a high-poly gun I've been making.

First off I did a simple bake without exploding, using rough settings just to make sure things were going in the right direction. Here's a snippet of what I got:

bake01.JPG

So then I exploded the model, used a custom cage (just the same model with a push and straightened up a little) and did another rough bake. This time it didn't go so well:

bake02.JPG

Now clearly something is going wrong with the way I've set out my UVs. Basically the whole model is only being mapped on one side with the other UVs sitting outside the main "box" by an offset of 1.

bakeProblem01.JPG

But what I don't understand is that nothing changed between bakes apart from exploding the model:

bakeProblem02.JPG

I'm using XoliulShader 2, non-Nitrous and have added that funky qualified line into Max .ini if anyone is wondering.

So like I said before, it's looks like it's something to do with the UVs being flipped? But I don't understand why it's a problem now I've exploded the model and not before? Also surely having flipped UVs is fine because that's how loads of models are done anyway?

I'm quite confused.

Replies

  • throttlekitty
    Options
    Offline / Send Message
    Looks like a flipped green channel?
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    Flipping/unflipping green channel on the XoliulShader doesn't resolve it.
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    Oh! This is interesting. I've just been testing the exploded bakes by themselves, without combining them in Photoshop, and they seem to be okay?

    I did try to normalise them in Photoshop...

    Okay so I've clearly fucked up somehow when putting together the maps in Photshop. I've tried normalising using the xNormal and NVIDIA plugins for Photoshop and they don't seem to fix it (the NVIDIA one makes it go crazy).
  • throttlekitty
    Options
    Offline / Send Message
    How are you going about combining them? For the most part, baked normals should be a simple exercise of select/delete.
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    That's pretty much all I'm doing. Combining via layers, deleting empty areas, setting the background to 128,128,255 and saving out a TGA. I found a thread on here about combining normal maps and will have a snoop through that.

    Okay back to having no clue again. This is really odd...
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    Okay so it's not even combining them in Photoshop, merely opening and saving again seem to make them have this result. Unless I'm using the original baked images I'm having trouble.

    I have no idea what is happening when I'm saving out of Photoshop. I've tried both TGA (at 16, 24 and 32) and JPEG now...
  • EarthQuake
    Options
    Offline / Send Message
    Wait, why would you need to combine different bakes? The whole point of doing an exploded bake like this is that you can bake the whole thing in one go.

    Your low should be one mesh, then just keyframe by element to do your exploded stuff.

    Now in addition to that, why do you have so many pieces split up? A more solid, air-test mesh would prevent needing so many individual mesh chunks, and give you much better UV usage, most likely.
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    Oh shit, I guess I don't need to combine do I. I originally exploded and baked separately so my PC didn't have to bake all that down in one go. But I guess it would be fine and is now more of a hassle than anything.

    And as for the low-poly, yeah... I made it trying to cut down on the poly count and now looking at it... yeah, it looks like there are too many separate pieces and too many UV shells. I'm going to continue with it for now to see if I can get a decent result. If not I'll have to just re-do a low poly and unwrap again. Chalk it up to a learning experience.

    But as for the original problem, any idea what was actually happening to make it go... "weird"?

    Thanks for the help man.
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    Oh now I remember why I split the mesh up. I wanted to export it to xNormal and wasn't sure how well xNormal would handle the full high-poly (being around 5 million).
  • AlecMoody
    Options
    Offline / Send Message
    AlecMoody ngon master
    if you have a decent video card xnormal will be fine with a lot more polys
  • throttlekitty
    Options
    Offline / Send Message
    Either way, saving a normal map as a .tga is kind of a serious issue if it's borking the normals. Could be color management, if you look at the re-saved normal map in another app, does the color appear correct?
  • Wesley
    Options
    Offline / Send Message
    Wesley polycounter lvl 13
    Cheers AlecMoody, it's good to know xNormal can handle more of that (I have a NVIDIA 560Ti with 2GB RAM so I should be okay).

    Also wow! Cheers for your suggestions throttlekitty, it turned out it was in fact Photoshop! Just tried Paint.Net and it worked fine. Also, it is way more automatic in terms of combining the normal maps (just gets rid of the waste straight away).

    I'm glad to know it was just Photoshop... I'm switching my mobo/reinstalling soon, so if it's playing up after that I'll take another look.

    Cheers again guys.
  • EarthQuake
    Options
    Offline / Send Message
    Either way, saving a normal map as a .tga is kind of a serious issue if it's borking the normals. Could be color management, if you look at the re-saved normal map in another app, does the color appear correct?

    Yeah a color profile or gamma issue, most likely.
Sign In or Register to comment.