/3/ - 3DCG

-Or UV layout Theory.

> select all polys
> auto unwrap them with 0 degree treshold
> put 0.01 for polygon islands distance
> press unwrap
Voila ! you got the most efficient UV space with absolutely zero stretching.

And anyone else find laying out UVs the slowest, most tedious part of the process?

>>383696
what program terminology are you using?
(What's the maya equivalent)

And if it's that easy, would it make a program like headus UVLAYOUT useless?

>>383698
Max. But max has a feature called UV padding which extend the UV bakes a few pixels outside without overlapping them on each other so you don't get the horrible UV seams. People usually tend to unwrap organic models seamlessly in a planar fashion, because UV padding didn't exist before, but programs like mudbox and max got around that little problem, so I don't see any point wasting hours or buying software spesifically to unwrap em.

>>383696
only useful if you're polypainting or don't mind seams at all.

>Try and do this with Blender's smart UV projection
>Computer explodes

I wish those fuckheads at the Blender foundation would stop working on their shitty cycles renderer that isn't even that good and go back to making Blender not something that came out of a 1940's taco food truck.

>>383697
Yeah it's certainly not the most exciting part. But to be honest I prefer unwrapping to weight painting. I dread that shit.

>>383696
if you are not into 3D for games, this is often the best solution, most VFX and movies can handle the extra render time created by having excessive UV seams and can also handle big ass textures needed when not doing tricks like overlapping UVs or using the tiling of the texture outside the 0 to 1 space
other than that, is just practice

We need to hurry up and make ptex standard. Game Developer Magazine already has examples of models using Ptex running in realtime game engines.

>>383923
ptex is useless for games. it takes too much processing power compared to the oldschool UVmap system. Ptex is only good for movie production

>>383923
You can just as well auto-UV each triangle individually (in Mudbox for example), same shit

>>383693
Export > obj
>import in zbrush, place on scene, press Edit
>use UV Master
> super quick uv making plugin with seams, symmetry, pelting and such..
tool>export>obj
>import into your modeler
> modify uv, move, stich, scale....


quality: above your normal uv crap, good, close to great
speed: hyper fast, minutes
difficulty: zero

use this for starters
http://www.youtube.com/watch?v=kOoOh5s8RE4


all is there.... simple, fast, efficient

>>383929
>>383925
>>383923
well can't there be a way of editing the UV seams of some areas that are dense with polys so that ptex doesn't make as many UVs?

>>383693
Keep islands proportional but maintain a certain minimum size.
Arrange the largest islands first.
Keep easy texturing in mind over efficiency. A "good" unwrap doesn't necessarily mean one that's good to texture, pic related. That also means keeping the orientation of the islands upright as often as possible, and clustering similar parts together.

The unwrap in your picture is a bad example. It has a ridiculous amount of duplicate islands that could be stacked.

>>383698
Maya has a good auto layout tool, it's in the UV texture editor. Use that. (but keep in mind that it won't stack identical parts, you have to leave dupes out and then stack them yourself)

>>383704
You have no idea what you're talking about. Every program can layout UVs with a certain distance, and pad the islands while baking. Seamless unwraps are still needed because UV cuts will fuck with your normals and will be visible at lower mips whether you like it or not, no matter how well do you hide them.

>>383956
I thought that was feels guy unwrapped..

good info jpg why can I never find uving info like this

>>383956
>but keep in mind that it won't stack identical parts

also keep in mind that you may want to make changes in the future where those parts may no longer be identical. in those cases it might be to your advantage to map them separately.
for example adding a tattoo to a character, doesn't need to be mirrored onto the other side of the body.

>>383800
they made many fixing in unwrapping tool lately. but there is only 1-2 guys on it. and his mom doesn't like him doing it. you maybe should spend some dollars on unwrapping tool projects so they can add some other people to it. (just saying but its great shit to program because you can play with different logarithms simple ways.)

personally i love blender unwrapping because being fucking crazy fast and working with seams. my only butt pain is that it doesn't have option to stretch or relax only into one axis directions. also it would be awesome to create better relaxing tools.

>>383956
I think the upper UV layout is very good.The lower one gives you texture stretchings in the face. Texturing organic models in a 2D painting program was yesterday. It's 2013, you paint your textures in 3d. (mudbox. mari, zbrush, blender, 3dcoat, bodypaint).

What do you guys think of Zbrush's Zremesher?

>>384099
God tier

>>384100
I really like it too. When I have some messy topology it really, REALLY helps clean things up, where as I can just fix up any other smaller problems by either adjusting the settings with Zremesher or doing a little bit of manual work. All in all it saves me hours of retopology.

OP Here,
So I unwrapped 3 models and merged them into one.
Now the 0-1 UV Space is fubar.

In maya, is there a way to auto LAYOUT (not unwrap) the shells you already unwrapped?

I made the UVs in 3d coat, and I know I can auto lay them out in 3d coat and a program like headus UV Layout, but is this possible in maya?

It'll save having to go back and forth between programs.

>>384270
Well crap, I found it.
Edit UVs>Layout>Option box.

>>383934
>>383934

mah nigga, this is what I do, I still don't remember where i learned it, and in my school everyone considers me as the UV God

>>384337
Where you learned pressing a button? Wow dude you're the king. Anybody who's not a moron knows that UV Master blows all other unwrappers out of the water. All the topology and voxel-related code they have in ZBrush pays off for this shit, it's basically free

OP Here again.

For the best results and maximizing use of UV space,
When should two smaller maps (like 1024) be used or be EXPECTED to give better results instead of using a single big map (like 2048).

I don't want to spend the time texturing something on a 2k map, to find out textures look more crisp if I used two 1k maps.

Is there any rule or theory to follow so that i know which approach to take?

>>384683
How the fuck would 2 smaller maps give better results than a large map that's twice the size of the 2 small maps. You goddamn imbecile

>>384683
Screen resolution. Determine how close the asset will be to camera. Will it take up at most 1k screen space? how much of the model is shown? If half the model is at 1k, then you would need two 1k texture maps at min. a 2k map would give you double the resolution you would need.

If i have time, i keep my textures at 8k. Then i'll downsample or make into exr tileable / mipmap. If it's some prop a 2 / 4k works. I'll often put similar props / multiple objects in the same sheet and work at 8k. Less files to keep up. Say i need to increase reflection i don't have to revisit each psd and save out a new file, just change the layer in one and save it out.

It also depends on your studio.

>>384697
>If i have time, i keep my textures at 8k

Jeepers.
Just for mapping multiple objects to it, or a single object or character with an 8k map?

>>384702
doesn't matter. I bake out all kinds of maps from zbrush and a vray dirt to use for masks in photoshop. If you sculpt a model, no reason to throw away the high frequency detail.

Example, i have stone work with small pores. I could use a bump map. or i could use the bump as a displacement in zbrush, and then bake out the surface data and use those maps in photoshop for the diffuse/reflect/reflectgloss / what ever else. And still use the original bump, or the normal map i've made.

With an 8k map, almost half the time i spend texturing is extracting surface maps. 4k is much faster.

This is a diffuse map at 4k, but in photoshop I worked in 8k. It's one map of three that i used for a jeep model. The nice thing about using all these layer makes is so that once i've made the diffuse map, i use levels to get the right reflect, refgloss, bump, w/e maps easily. It's not a right or wrong way, it's a system i've come up that works best for me. Some people like to take photos, warp them, blend them, or use them for maps. If you dont have a lot of time and you want to get a lot of detail into your models it works great. I just like to spend my time sculpting the detail.

Is it a bad idea to overlay multiple surfaces in the same space if the objects are all the same?

For example, a model that had tons of rivets or screws. Just placing all those faces on the same area of the map.

>>384787
No, that's a very valid way of saving space on a texture sheet.
If you were making anything for games, you always end up trying to figure out how many things you can stack up and re-use.

Really noob question(s) here but since it's a UV thread, i figure it's the best way to ask.

1) How do I know what size to make the texture sheet? I see a lot of UV's where everything is unwrapped and the sheet holds everything just right. Do I just unwrap, move the verts myself and tak up all the space I need?

2) For stacking, if an island is not the same size, but same part, could I resize it and stack on to another similar piece? There wont be any crazy stretching, will there?

3) Is it better to have a lot of UV parts (islands) or just a minimal amount? I"m sure this is just me doing shitty texturing, but I find it hard getting the same colors on all the pieces when there's so many of them. I guess that might just require some more practice on the photo editing part.


>>382960
Here's a picture of my first attempt at texturing a model with 'details' like wrinkles and all that. And here's a picture of the uv soyou guys can see how bad it is and possibly give some tips.

>>384888
>How do I know what size to make the texture sheet?

There is a term called texel density/texel resolution that you should be aware of. First off you need to know what a texel is; a texel is the term for a pixel on a texture. Imagine you have a white texture with a single black pixel on it as you view that black dot in the video game you would refer to it as a texel. The reason for this is because as you shove your nose up into that texture in the game world that black dot could cover far more than one screen pixel so you can no longer really call the black dot a pixel so to avoid confusion it is called a texel or texture pixel. Now imagine you apply that white texture with a black dot to both a small dice and a large building it will look far more stretched out and blurry on the building because it is covering far more surface area, its texel resolution is far lower.

So with that explained the way you judge how large to make your texture sheet is based on the texel resolution of the game world you will be importing it into. If you can apply a checker box pattern to an object from that game then you can quickly gauge the texel resolution and if not just go with your best guess.

>Do I just unwrap, move the verts myself and tak up all the space I need?

Packing UV islands to have as little wasted space as possible is a skill you develop as you practice uv mapping. You learn how to unwrap things so that they are not weirdly shaped and hard to pack. It is mostly done by hand if you want the best result you can get. The reason you want to have as little wasted space as possible is because that wasted texture space translates into wasted video memory on the video card and in system RAM. It is a real issue to have too much wasted space because memory limits are something that developers bang their heads up against all the time, so don't be wasteful.

>could I resize it and stack on to another similar piece

This depends on the video game engine, some handle it better than others. Most all modern engines handle it well and the few older ones that don't mostly have issue because of the normal map texture. So you should try it and see and chances are everything will look fine.

> There wont be any crazy stretching, will there?

This comes back to the texel resolution we were talking about a bit ago. If you scale a uv island it will change its texel resolution but there is always some give you have with that so once more just try it and see if everything looks fine.

>Is it better to have a lot of UV parts (islands) or just a minimal amount?

You want to have as few uv islands as possible because everywhere you add a seam is a place where you have just added a hidden vertex. You don't know it is there, and it is not reflected in your polycount but the cost to render your model has just gone up. Here is a great tutorial that will help you understand this concept better, it is a must read.

http://www.ericchadwick.com/examples/provost/byf1.html

>>384894
I just noticed I linked the wrong part of the tutorial, the relevent bit for you is part two linked at the bottom of the page of the link I posted. It would be a good thing if you read the first part though.

>>384892
>>384894
>>384895
You have no idea how grateful I am for your time and information. That's a lot of information I was really lacking but thanks to you, I'm a step closer to getting better at this 3D.
I'm reading both part 1 and 2 since I'm pretty sure I can use all the knowledge it has.

Again, thanks for all your help. After writing down more notes and studying some more, I'm going to give the dinosaur texture another try. Hopefully you'll be around when I put it up.

Thanks again

>>384916
It does get easier with time.

One method I use, as I unwrap I keep all my finished parts off to the side, when every part is unwrapped, I arrange them from left to right in terms of how much detail that part will need, and I pack those parts first. (face, high detail/exposure areas. Hand details. Shit like that).

So happy this thread is still here.

Please review and tell me how to make my UVs better next time.
Anyone.
:s

The Rocket Launcher looks fine from afar, but up close = blurry, bad texel rez etc.

This was unwrapped in 3d coat. I love it cause as soon as you unwrap everything, it autopacks everything to scale relative to each other.
I guess I could have enlarged some parts myself, but I really didn't want to fuck anything up.
Looks like I did anyway.

>>388101
> I love it cause as soon as you unwrap everything, it autopacks everything to scale relative to each other.
I guess I could have enlarged some parts myself, but I really didn't want to fuck anything up.

Unwrapping things efficiently comes with a bit of practice, it's nice having the shells sized relatively to each other but you have so much wasted texture space there that it's kind of all for naught.
You want to make sure the main parts are filling up as much of the UV space as possible and squeeze the lesser important parts around them. It's not too hard to judge by eye how big things need to be.
I think if you gave unwrapping and laying out the UV's of the rocket launcher another go and try to fit things together tighter and not waste so much space you will get better resolution.

>>388101
> I love it cause as soon as you unwrap everything, it autopacks everything to scale relative to each other.
Umm, what the hell exactly were you using before? Because pretty much every modern 3D software does this. Maya, Max, ZBrush, Softimage...

>>388101
Looks okay.

One thing you might want to start practising is overlaying UVs for similar objects. If you overlapped every duplicate object, and you live in a perfect world, you can then double the size of your UVs giving you more resolution to work with at the expense of texture repetition. I've highlighted in different colours a bunch of very similar objects I can see with little texture variation. The balance here is if these sections are going to be seen at the same time, and if the repeated texture will be noticeable.

Another thing you can do is cut symmetrical objects in half, and mirror their UVs. I've marked some of these sections with a dotted yellow line. I'm a bit tired and haven't thought it through 100%, so that red lines you have going may not work with the symmetry. Again, this comes down to whether or not you'll notice if the textures are mirrored across the surface.

Using these techniques will allow you to 'increase' the apparent resolution of you texture (less blurry pixels) without actually increasing the texture size.

Don't be afraid to add a few extra seams if it can reduce wasted space on your UVW map. Notice the areas in yellow, these can be broke off to make it possible to better pack the uv islands.

The red circles point out issues with your islands being too close together. Video cards have limited amounts of RAM so video game engines use mip maps to save space in RAM. A mip map is a lower resolution version of the texture that is generated automatically from the texture. If your islands are too close together then when the lower resolution mip maps, there can be several, are generated the islands will share the same pixel and can make seams look awful. You want around a 6 pixel gap on 1024 textures and 8 pixels on a 2048 map. When it comes time to bake your texture look for an option called padding and enable it (3d-coat does have this ability). Padding will cause the color of the edge pixel of the island to be pushed out into that empty gap so that when the mip-maps are generated the color of the seam pixel does not change.

Also, it is hard to tell from this image alone but I would suspect that you can stitch several of these seams back together to make larger islands.

>>388180
Dude, you are the god damn man.

Now you are the UV God.

>>388180
>Don't be afraid to add a few extra seams if it can reduce wasted space on your UVW map.

yeah, but then won't I run the risk of having more visible seams on the final model?
-Especially if using a photosource texture (photos of metal, rock etc.)

>>389177
Not wasting uv space is priority number one, reducing seem count comes after that because video games tend to be gpu bound not cpu bound.

Also there are techniques to reduce how visible seems are:

-After you create your uvw map and texture make a clone of your model
-Create a new uvw map for this clone with all the seams in different spots from where they were. Do not worry about how many seems you have on this version or how much wasted space you have.
-In x-normal or any other texture baking program set your actual model to the high poly slot and your clone model to the low poly slot and bake a diffuse/color map. (make sure the resolution of the texture you bake is high enough that your texel density does not drop)
-This new diffuse map will now have your visible seems in the middle of the clones uv island and you can now paint them out of existence.
-Once this is done set your clone model to the high poly slot and your actual model to the low poly model to the low poly slot and bake a diffuse map to transfer everything back.

>>389211
So you're saying.....

1 Make final model and texture it

2 Then make a clone of that model, but with different UVs...any kind of auto-unwrap should work, right?

3 Transfer the texture via baking from the first model to the new clone

4 paint away seams

5 transfer from new clone back to original model.

If that's correct, to you have a tutorial of that technique? Or video of it?
-And any other "techniques to reduce how visible seems are"?

>>389215
I did a quick google search before making the post trying to find a tutorial and failed so I made the post instead. I am sure there is a tutorial out there somewhere if you can think of the proper incantation to chant to the all mighty google.

As for other techniques the main on is just being logical about it and putting seems in places that will be hard for the end user to get a view of. For example on a weapon it is hard to get a good view of the bottom of it so put seems there.

>>384872
Unless you plan to bake AO or other lighting stuff.

Perhaps you could make a separate UV unwrap and have a half resolution lightmap. But does the cost outweigh the benefit from repeated textures?

>>383801

this.

Unwrapping is relaxing, it's like creating a papercraft layout from an already completed design.

Here's what I'm gonna unwrap and transfer normals.

I was about to:
cut the low poly in half,
transfer normals
then create diffuse,
But I forgot that I don't want the diffuse job to be symmetrical, So I'm gonna have to unwrap the whole thing (less texel density).

Still, hopefully texel resolution ends up better than the gun.

>>389834
I think it depends on the program you use.
3D coat is great for retopology, but (to my knowledge) you can't move the pieces around then unwrap them seperately like maya, UV layout etc.

Unwrapping in UVlayout I think is the best.
Expensive as hell program, but I think it's the best at what it does.

>>390410
Literally looks like what I flushed down the toilet this morning

>>390411
>you can't move the pieces around then unwrap them seperately

I am not sure what you mean by this. 3d-Coat lets you retopologize in layers if you don't want all of the mesh to be one object if that is what you mean.

>>390410
>But I forgot that I don't want the diffuse job to be symmetrical

Some engines let you use different uvw sets for diffuse and normal to side step this, I think the unreal engine allows it. Anyway with next gen games just around the corner and both major consoles using 8gb of shared video memory having a fully unique texture is probably the correct direction anyway.

Also work on your topology, your low poly model does NOT need to be all quads. Adding a few tris here and there would allow you to remove a bunch of redundant geo.

>>390457
>3d-Coat lets you retopologize in layers
Maybe that's what I'm looking for. Sound like it.


> Adding a few tris here and there
It has them.

Transferring normals in xnormal.
Got this problem.

:s

I'm thinking of just brushing out the area with a soft brush and hat normal map blue color.

>>390610
forgot pic

>>390610
if you calculate the distance in xnomal, it should get rid of it somewhat. if it does but still has some weird artifacts going on, you can layer the new render on top of the old one so it blanks the artifact out.


or you could use a cage, but this is coming from an amateur so there could be better ways to do it.

>>390611
Is there no convenient recalculate normals button?

>>390611
Yes, you can just paint out artifacts like these. Don't be afraid to edit normal maps manually.

However, you might want to play with the ray distance settings first to minimize the areas that need touching up. Make sure backface culling is on.

>>390633
>recalculate normals
>>390665
ray distance settings

This is what I really need guys,
Is there an intermediate/advanced definitive retopology guide/tutorial?
Everything I find is just making a simple cage, UV wrapping, then baking, but for different programs (blender, zbrush, 3d coat etc.)

I really need to understand why my bakes come out fubar instead of it being a roll of the dice whenever I get results of my bake.
It's infuriating thinking it's just a case of remaking the entire cage, then unwrapping it, then hoping for the best results.

>>390633
There's a calculate normals button that gives you some numbers, but I don't know where/how to apply them.

I'm guessing this is where I'm supposed to used that "hand plane" program.

This thread is awesome, thank you very much.

>>390633
finally found out how to work it.
I thought I had to "apply" the calculations somehow, but once calculations are complete, it auto applies or something.

Anyway, here are my final UVs, auto arranged in UVlayout (cause seriously, fuck laying them out by hand).

Face UV shell was enlarged a bit for texel resolution (in theory, they'll be looking at the face a lot, amirite?)

Now I can start texturing.

now rigging....

>>392507
> overlapped UV islands
gurl iz u 4 reel

>>392864
IKR
It's one of those things you don't notice when it's done, but when you put it out there for the world to see.

It's fixed in the final maps:
>>392863

Until next time.

-_-7