Jump to content

UV-mapping problem


thePhilanthropy

Recommended Posts

I'm using 3ds max 8. I'm currently working on a rather big mesh (82000 vertices). The UV map is already finished. But when I try to render it I can only select 4096px as a maximum resolution. For smaller models I can go up to a resolution of 10000px.

I have no idea why the program won't let me render it higher. I have pretty high ram (>3gb), so that shouldn't be the problem. But I could be wrong, of course.

 

Does anyone know a trick for this? Or an alternative program that definitely can do it?

 

Thank you in advance!

 

 

PS: yes, 4096 is already pretty big, but this model is extremely complex. Below 4096, texturing is almost impossible. And 8192 would be way better for that. Doesn't mean the texture will be that big, but it would certainly help.

 

PPS: no, nifskope is not an option. I already tried.

Link to comment
Share on other sites

I'm using 3ds max 8. I'm currently working on a rather big mesh (82000 vertices). The UV map is already finished. But when I try to render it I can only select 4096px as a maximum resolution. For smaller models I can go up to a resolution of 10000px.

I have no idea why the program won't let me render it higher. I have pretty high ram (>3gb), so that shouldn't be the problem. But I could be wrong, of course.

 

Does anyone know a trick for this? Or an alternative program that definitely can do it?

 

Thank you in advance!

 

 

PS: yes, 4096 is already pretty big, but this model is extremely complex. Below 4096, texturing is almost impossible. And 8192 would be way better for that. Doesn't mean the texture will be that big, but it would certainly help.

 

PPS: no, nifskope is not an option. I already tried.

Well, most videocards can't handle textures larger than 4096x, and older ones can't handle larger than 2048x. So exporting a UV template larger than that is likely not that useful. Technically speaking, having 2-3 2048x textures is far less straining on a videocard than a 4096x texture, and since there are still some who cannot even manage a 4096x texture, having 4 or 5 2048x textures is usually preferrable. Even if you happen to have a beast of a videocard, and can manage higher textures, chances are that anyone using that mod won't.

 

When I make UV maps, I usually just export at the usual 512x, and just enlarge the exported map to whatever size I am working in (usually 1024x) in a paint program. Realistically, the uv map should only be there as a guide for texturing, and trying to do anything too precise using the UV map usually ends up having issues due to the real texturing being slightly larger than what is rendered for the template.

 

Since you're using 3dsmax, you can rather easily take parts of a mesh, divide them from the rest (retaining their mapping), apply a new unwrap modifier to that separated group, edit the uvw, and move things around to be used as a new texture. Once you've applied a different material (texture) to that group, you can even re-attach it to the main mesh, weld any verticies that might have been split, and keep those two parts using different textures (using a material with multi is another option if you know how to do it). There really is no reason to use a single texture that large for anything that isn't a single face.

 

As for the confusion, I think you might be mixing up render types. Normal image renders can be rather large, uv renders however have to be small enough to be handled for the purposes of texturing, which I believe (and am probably wrong) is one of those limitations based on the videocard and what version of directx you are using.

 

The count that matters is not verticies, but faces. Anything over 15k faces had better be a static, and anything over 30k faces should likely be broken into more pieces.

Link to comment
Share on other sites

Thank you for the elobarate answer. I already knew all that, no disrespect.

My question was rather directed into another (more theoretical) direction. I do know that it would be better to split the mesh (+UV/texture) into several parts (which I will do eventually, anyway). But, what I really want to know is why 3ds max doesn't allow it in the first place.

I can easily imagine an algorithm that works recursive and doesn't use much ram (or vram) at all, no matter what resolution.

And no, I'm not mixing up render types.

I would just be interested in a method to render BIG UV-templates.

If you want a model with the least possible seams a big UV might be the only solution. SO, I'm wondering, how do the guys in the industry do it?

 

PS: about vertices/polygons/faces: Of course, what matters in the end are faces. But that number is actually rather unclear before any export. In 3dsMax I have 82k vertices with aprox. 150k faces. But the face-count changes with the format I export to (while the vertex-count stays basically the same). That's why I'm refering to that number.

Link to comment
Share on other sites

Well, some of what I was trying to explain is that ultimately the size of the uvwmap becomes irrelevant since the actual locations of those textures along the seams are always off by about 0.5% of the end texture size, so a 512x texture will have actual seams about 2-3 pixels out from the rendered "edge", a 1024x would be 4-5 pixels, and so on as the template gets larger. What really matters is the actual texture, and I would image that there is always quite a bit of back and forth work between the 3d program and the texturing program to really tweak things right. On a very complex model, with many pieces, they almost always use multiple textures (look at the IC palace some time. There are many textures which are used only there. Just like anywhere else, it usually makes sense to do the repetitive parts as their own individual textures/elements, and count on being able to break up the remaining parts into different textures instead of trying to cram it all on a single one. So far as I've seen from all of my time modding, the only place where I saw any sort of large scale texturing like what I think you are talking about was with houses for Sims 2 (houses were disected from 4 directions with textures being generated based on what walls were seen from that direction, and grouped by "levels"), and that was only for the LOD versions. Most often, they just try to cram as much as they can on a single texture, and try their best to leave a little bit of space between elements. Should take a look at how some of the weapon textures were arranged. The daedric longsword texture is actually divided into two pieces in the middle of the blade, but you never see a seam. This is handled not with some large uv map, but clever layout and good texturing skills. Many of the weapons make use of these sorts of tactics.

 

For anything cinematic wise, I would imagine it's the same deal. Huge textures would only add to rendering time, so as much as can be done with smaller, tight textures, usually the better. The only change is that they have a higher paid staff, and can usually tweak things fine post production. Most cinematic models I've seen also seem to borrow quite a bit from material properties, with several channels and sub materials, and not merely texturing.

 

The point I am trying to make here is that the uv template is just a template, and not meant to be an end-all pattern for any texture work. It exists just so that when you do start on a texture, you have some idea where pieces are and how they are oriented. The final alignment stuff can all be done either on the texture or adjusting verticies. For an extremely complex project where you are filling up most of the space for the texture, it can usually help to render separate templates for groups of parts instead of trying to put them all on a single template and make sense of it. Rendering it as a filled, or normalized template without visable seams, overlap, or faces can also make things a bit less confusing. But really, the template is only useful until you have something sized to the actual texture to start aligning everything to. Creation of the UV template should be more focused on just getting things oriented, unwrapped, and separated into meaningful clusters, than trying to figure out exactly where every point will go, and trying to make every seam match up perfectly. A gap between elements is often much easier to work with than trying to match them up along an edge (since unwelded might end up overlapping that 0.5% I mentioned earlier). There really isn't any reason why you would ever need a uv template that large because the template is only a guide, and anything after that is really dependent on forgiving the mesher set it up to be, and how skilled the artist is in blending the seams.

 

Just for the heck of it, I tried rendering a 8000x8000 template, and ended up freezing before anything could be saved. So either my videocard is crapping out and Max 8 allows for templates that large, or it errors out because the program cannot export that large.

 

One of the best things you can do if you are having trouble getting a handle of UV mapping is to try mapping new objects to vanilla textures. It helps you move beyond the whole "planar" approach, or getting obsessive about seams. Instead, it focuses more on moving vertices around on the texture to try and match things up right. I'm not suggesting that you may not be experienced, but instead am offering a different approach that you might not have considered where you are essentially working backward from a finished texture, and building the uv around that. Nearly everything I make uses this approach at some point just to get a quick idea as to how faces might be arranged.

 

I still can't think of any instance where an extremely large uv map would be particularly useful, as opposed to just scaling a smaller one up to the size of the texture, but will think about it more.

Link to comment
Share on other sites

I would give you more kudos if only I could.

But while the alternative approach you offered might be insightful for simple meshes(e.e. a sword), it is not very applicable for complex ones (e.g. creatures).

I'm just worried that MAX will restrict the template even more when the mesh gets bigger (in theory). But I just tested it with a mesh that has 240k vertices and it still allows 4096, so that doesn't seem to be the case.

 

I'm not a pro with max. But I'm not a n00b either. Actually I'm pretty good, but I'm also always interested in new tricks (regarding the theory, or this particular program).

Link to comment
Share on other sites

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...