Home Marmoset

GroupID and other baking issues (3.04)

polycounter lvl 14
Offline / Send Message
Monophobe polycounter lvl 14
Hey,
Apologies if these issues have been covered elsewhere, but I figured this would be a good place to report some issues I've come up against in 3.04 (and maybe find some workarounds/answers).

It appears that any change to the FBX that contains my high and low poly (all named for baking groups), results in the colours in a GroupID map changing when baked. I realised I needed to fix up something in my high poly last night, which I'd already done a lot of work on in Substance Painter, but it was a significant enough change that it warranted rebaking the GroupID. I had no new meshes, just edits to the existing ones, and all names remained the same. This colour change led to a good hour or so in Substance fixing up all my colour selection masks.

I wondered if maybe it was because I'd changed the high res, but I just tested removing a single poly from my low and once again the colours were randomly assigned.

It makes sense to me that this could be handled by just assigning a colour to each high res mesh the first time it's loaded into Toolbag, and storing that colour along with the mesh name. Ideally this would also be exposed in case we ever need to tinker with the colours.

I appreciate I could assign vert colour to my high res instead and bake that map out, but that can be fairly time consuming on complex models, and somewhat defeats the purpose of the GroupID bake.


Also, whilst I'm incredibly impressed with the Toolbag baker, I do consistently have an issue with any meshes I've altered the offset and skew on. Each time I load my saved project, these meshes bake as though the adjustments haven't been made. I have to select each modified mesh (thankfully only a couple in my current project), make some kind of edit to either the offset or skew, and then undo it. After that it's like Toolbag wakes up and remembers there had been edits made in the past.

Replies

  • EarthQuake
    Hey,

    Firstly, the material ID colors are automatically generated based on the number of materials you have applied to your high poly mesh. If the number of materials does not change, these colors shouldn't update, so that may be some sort of bug. I'll investigate that.

    If you have specific colors that you would like to use, you can define them with the high poly materials directly, for instance, apply the color to the albedo, and then bake an albedo map. If you need the albedo slot free, you can apply the colors to the emissive slot.

    We're aware of the skew/offset map issue and that should be fixed for the next release. Basically the changes are being saved but they do not initialize until you select the low object, so yeah selecting those when you open an existing file should sort it out.
  • EarthQuake
    Alright I did a quick test here.

    1. Editing the low poly mesh (moving faces around) did not change the MatID colors
    2. Editing the high poly mesh (moving faces around) did not change the MatID colors
    3. Editing the high poly materials (adding a new material) did change the MatID colors

    So #3 is expected and working as designed, if additional materials are added the colors are regenerated. How the system works is that it will pick colors that are as mathematically different as possible, so when new materials are added that calculation happens again.

    If you're seeing the MatID colors change in case #1 or 2, or if you have another case where you can get this to reliably happen please let me know.

    There may be some cases where mesh edits result in new materials being created or applied to your mesh(es) by your 3D app, which would bring us back to #3, so make sure that is not the case.
  • Monophobe
    Offline / Send Message
    Monophobe polycounter lvl 14
    Apologies, I just realised I wrote GroupID when I meant ObjectID. I can't be 100% sure I didn't inadvertently end up with another material, but my high poly only had 1 material on all objects. I was attaching new mesh to existing objects in max so it should have just used the existing one.

    When I get chance I'll do some more tests - there's every chance this is a simple mistake on my part. Thanks for your time.
  • EarthQuake
    Monophobe said:
    Apologies, I just realised I wrote GroupID when I meant ObjectID. I can't be 100% sure I didn't inadvertently end up with another material, but my high poly only had 1 material on all objects. I was attaching new mesh to existing objects in max so it should have just used the existing one.

    When I get chance I'll do some more tests - there's every chance this is a simple mistake on my part. Thanks for your time.
    Thanks for the update, I misread and thought you meant material ID maps. Sorry about that, I've repeated the same test here and have found the bug that you're reporting. We'll look into fixing this for the next build.

    In the mean time I would suggest using a unique material per high poly object and baking a material ID map, this should be more stable.
Sign In or Register to comment.