Home Technical Talk

General Question about modeling simple props with many shapes

polycounter lvl 12
Offline / Send Message
amandell polycounter lvl 12
Is there a rule of thumb to keep in mind while modeling scenery props that have a lot of shapes that when boiled down are basically just 3D primitives?

For example, if you were building wooden steps that are separated and a hand rail, the whole model could be made up of elongated cubes and cylinders. Floating pieces that are placed accurately.

Is this a bad habit to get into? Are there more correct ways of handling objects that (for ease sake) could be built with all floating shapes?

Replies

  • Bek
    Offline / Send Message
    Bek interpolator
    If you're talking about floating geo specifically, it depends. If you're going to be baking a normal map from a high to low, there are some 'extra' things to consider. Have a look at this thread: http://www.polycount.com/forum/showthread.php?t=117017
  • sprunghunt
    Offline / Send Message
    sprunghunt polycounter
    well if you were being conservative with textures you wouldn't build the whole set of steps. You'd only build a few risers and a straight section of handrail and some other pieces. Then you'd construct your steps using these bits by repeating the geometry.
  • amandell
    Offline / Send Message
    amandell polycounter lvl 12
    What I'm getting at is, where do you draw the line (when making lower poly models) where tacking geometry together without welding vertices becomes a bad practice?

    For instance, let's say you were developing a game character that had many accessories to the character's outfit (belts, extra gear, water canteen, etc.) but these accessories were never used/animated off of their original placement on the model. Should meshes like those be left as floating pieces, or should they be attached as one mesh?
  • tynew
    Offline / Send Message
    tynew polycounter lvl 9
    There is no harm in leaving those accessories as floating I'd say. It would be a pain to model those all into once piece with the character. Unless your talking about subobjs, then yeah you'd want to attach them so they can share UV space.
  • cryrid
    Offline / Send Message
    cryrid interpolator
    amandell wrote: »
    What I'm getting at is, where do you draw the line (when making lower poly models) where tacking geometry together without welding vertices becomes a bad practice?

    For instance, let's say you were developing a game character that had many accessories to the character's outfit (belts, extra gear, water canteen, etc.) but these accessories were never used/animated off of their original placement on the model. Should meshes like those be left as floating pieces, or should they be attached as one mesh?
    Belts you may want to model attached, or have them created via the texture. Then they won't clip through the model entirely. On the otherhand, if a larger object clips, then it at least won't completely disappear.
  • Gestalt
    Offline / Send Message
    Gestalt polycounter lvl 11
    I'd say if elements don't have a smooth transition it's almost always better for them to be separate pieces. There is nothing wrong with floating geometry besides maybe having an AO baking error or something like that.

    Sometimes it's easier to just bake things down into your low poly. It's really a case by case thing. It depends how obtrusive the thing is. For pieces that are more modular I would say go for floating geometry.

    For one you're able to keep the geometry and normals of your pieces cleaner because you don't have to join objects. You can also usually get away with fewer verts for your low poly for the same reason. When you have separate pieces of geometry they conveniently make their own UV islands (which you might have done anyway), and you can add and subtract elements much more easily. It's much easier to iterate or make multiple versions of the same piece when the geometry and UVs are unique to the part you want to adjust.

    The downside is your pieces won't have a transition, but for many things that's the case in real life anyway and you could always make additional floating pieces to cover the seam or join the meshes (like hinges and other couplings).
  • pablohotsauce
    Offline / Send Message
    pablohotsauce polycounter lvl 7
    Minor thread hijack here, sorry. I was looking at the thread linked above (http://www.polycount.com/forum/showthread.php?t=117017) in which EarthQuake gives advice on this very thing, and I was wondering something but I didn't want to necro that one.

    EarthQuake says that one of the benefits of having large continuous lo-res meshes is that you end up having correspondingly large and continuous UV islands, which are easier to paint, waste less texture space, etc. But, wouldn't you have to break these large UV islands up into smaller islands regardless, in order to A) be able to unwrap the continuous 3D meshes in the first place (not every big mesh is amenable to being unwrapped as one giant island), and to B) account for smoothing group splits (which, on a hard and angular object like a gun, you'd have tons of)? So wouldn't you end up w/ tons of small UV islands anyway? Or does this all depend on the complexity of the mesh, and whether it's organic/inorganic?

    @Gestalt: but in the same thread above, EarthQuake mentions that intersecting geometry in low poly creates aliased edges where these separate pieces meet. Is that a big problem or no?
  • Gestalt
    Offline / Send Message
    Gestalt polycounter lvl 11
    I'd say it really depends on your needs.

    If the part is independent of the rest of the mesh and it protrudes I would probably keep it as its own piece.

    Having it as its own piece with its own UV island will let you assign a material to it better, it will give you cleaner separation of you materials, having it as separate geometry will let you adjust or remove the piece very easily if you wanted to assemble things more procedurally, it will let you replicate the piece easily, it will keep your low poly mesh much cleaner and the normals much more accurate in many cases.

    Generally I haven't seen much of an issue visually with independent meshes, although you'll definitely get a better transition if everything is baked down as one piece due to the nature of doing things that way. It's really your call in the end.
  • EarthQuake
    Yeah it really depends, if we're talking about completely separate objects that you could repurpose in different ways, that should always be modeled as separate mesh chunks.

    If we're talking about unique elements of one asset that will only ever be aranged in one way, you're generally better off making the low as water tight as possible, with only separate chunks for areas that need to be removed(optional accessories), animate, or will otherwise cause bake errors due to complex intersections.

    ar02wiresbig.jpg
    Heres an example of the highpoly for a FPV weapon model, very many separate mesh chunks

    3PointStudios_Brink_Maya_ar02_04.jpg

    And the lowpoly, most of the main mesh is combined into one object (except the rails which were kept separate to keep tricount down at client request, I normally wouldn't model rails like that).
  • tynew
    Offline / Send Message
    tynew polycounter lvl 9
    In regards to that being a single object Earthquake, how is it possible that you have a clean baking cage? Even with such a tight low poly, If I had a similar mesh in max/xnormal and pushed my projection/bake cage out, wouldn't there be a ton of intersecting areas especially where there are a lot of tight angles going on?

    You even modeled the sights into the main section/picatinny rails? If that were me my cage would go ballistic around that area:poly122: Maybe I'm just outdated with baking, I still explode each individual section to achieve clean bakes, and that requires them to not be modeled into the main mesh.
  • EarthQuake
    The rails, sight, bolt, bolt flap, charging handle, and mag are separate bits as they all(except the rails, which were done to client spec) can be removed/animated. The main bulk of the body/stock are all one mesh though.

    Not sure why you would get bake errors from having the lowpoly as less mesh chunks, generally the more chunks you have the more you need to explode/deal with to avoid intersection errors and the like. So its actually much easier to get simple, clean bakes.
  • Gestalt
    Offline / Send Message
    Gestalt polycounter lvl 11
    Yeah that's another part of the equation. If you have separate meshes the expectation is that you're baking the details with an exploded mesh (treating the elements as separate). If you can cover an area with a simplified low-poly mesh and just bake the high-poly to that then that's probably fine in most cases.

    I think the mentality for individual elements is especially geared towards the high-poly, much more than the baked low-poly. For the high-poly the more you can break up the more you can adjust things and get rid of things.

    A fast way to work that way in Zbrush is to use Extract. You see people use it a lot for their hardsurface work.
    http://youtu.be/mCgLNfuX47M?t=6m33s
    You mask off where you want to extract a plate or a screw or whatever and it gives you a new subtool that is flush with the mesh. It's much nicer to have details as separate things than to keep them all directly as part of the same base mesh.
  • pablohotsauce
    Offline / Send Message
    pablohotsauce polycounter lvl 7
    Thanks for the replies. A couple things:

    1. EarthQuake: I wanted to clarify my original question above. I was just being nitpicky about something you said in the older thread, about one of the benefits of having a continuous lo-poly mesh being that you could have a continuous UV island, which is easier to paint. But, as I understand it, you should create UV splits wherever you have smoothing group splits on your lo-poly. And, in that case, you'd end up w/ a bunch of smaller UV islands anyway, as opposed to large ones. I took a section of your gun model and painted over some theoretical smoothing splits and corresponding UV splits. Am I wrong here about having to split UVs along smoothing group splits? Or are you not worrying about smoothing groups here or something? I'm just going off what I've read in other threads on here, and what I've seen Alec Moody and some other people talk about.


    eq-smoothing-01.jpg~original

    eq-smoothing-02.jpg~original

    2. On a different subject: in your guys's experience, when baking things separately, does the method of assigning matching material IDs between the lo-poly and hi-poly meshes and then baking by matching IDs generate the same quality bake that you'd get by manually exploding the meshes instead? Or is exploding still better somehow? In my testing I seem to get the same results between the two methods? So why would people still bother manually exploding, vs. assigning mat IDs?

    3. EarthQuake again: off-topic, but I spent most of yesterday reading through your article on Marmoset's site about PBR in practice and taking notes like crazy and studying the screens in the article, etc, so it's an honor. :) It's cool that you post here and take the time to help us noobs out.
  • Bek
    Offline / Send Message
    Bek interpolator
    1) You're not wrong— a hard edge requires a UV seam but a UV seam does NOT require hard edge—it is almost always beneficial though. (When talking about a model that will use a baked normal map)

    2) Not all applications support material IDs (or cage baking for that matter)
  • tynew
    Offline / Send Message
    tynew polycounter lvl 9
    @pableaux
    With point 2, I watched Xoulil's High poly bulldozer tut. He mentioned that mat ID baking is more useful when you have something with a ton of objects, in comparison to a simple weapon. I haven't had a chance to try it out yet, which method was faster for you?
  • EarthQuake
    pableaux wrote: »
    Thanks for the replies. A couple things:

    1. EarthQuake: I wanted to clarify my original question above. I was just being nitpicky about something you said in the older thread, about one of the benefits of having a continuous lo-poly mesh being that you could have a continuous UV island, which is easier to paint. But, as I understand it, you should create UV splits wherever you have smoothing group splits on your lo-poly. And, in that case, you'd end up w/ a bunch of smaller UV islands anyway, as opposed to large ones. I took a section of your gun model and painted over some theoretical smoothing splits and corresponding UV splits. Am I wrong here about having to split UVs along smoothing group splits? Or are you not worrying about smoothing groups here or something? I'm just going off what I've read in other threads on here, and what I've seen Alec Moody and some other people talk about.

    Yeah, so of course the uvs are going to be split in some places, and since hard edges are free at uv spits I used them as well, the point isn't to have 1 huge uv island for your mesh, as that is unrealistic and wasteful(much harder to pack the content efficiently).

    However, the very important thing to remember is uv splits or hard edges don't cause aliasing when your mesh is one solid, continuous chunk. As long as your bake is set up correctly, and your normal map pipeline is synced, you won't see any seams or aliasing there. On the other hand, you will always see aliasing when one object intersects into another (unless you have very high anti-aliasing, and even then it doesn't look as nice as you won't get a bevel at the intersection point).

    In general, it doesn't even make much of a difference in triangle/vert count to do this, and you can squeeze more texture resolution out of your maps (less space wasted representing areas you can't see). It also, again, makes it simpler to set up your bakes.
    2. On a different subject: in your guys's experience, when baking things separately, does the method of assigning matching material IDs between the lo-poly and hi-poly meshes and then baking by matching IDs generate the same quality bake that you'd get by manually exploding the meshes instead? Or is exploding still better somehow? In my testing I seem to get the same results between the two methods? So why would people still bother manually exploding, vs. assigning mat IDs?
    I don't do this personally, I do the exploding thing. The mat ID thing doesn't work in Maya afaik (or maybe they added it in a new version?). Setting up explode bakes shouldn't take more than five minutes or so, so its not a bottleneck in any way for me.
    3. EarthQuake again: off-topic, but I spent most of yesterday reading through your article on Marmoset's site about PBR in practice and taking notes like crazy and studying the screens in the article, etc, so it's an honor. :) It's cool that you post here and take the time to help us noobs out.
    Awesome, glad to hear it!
Sign In or Register to comment.