Page 3 of 3

Re: MakeClothes Universal Fit -- How I Got Pretty Close

PostPosted: Tue Jan 23, 2018 12:28 pm
by wolgade
MargaretToigo wrote:Vertices assigned to *groups with three points of reference do behave differently from those that have only one point of reference,

All vertices of a proxy have three points of reference on the human mesh. In other words: Every proxy vertex gets influnced by a triangle on the human mesh. The difference between the regular fitting and rigid being that regular proxy vertices get individual triangles, rigid proxy vertices share the same triangle.

Re: MakeClothes Universal Fit -- How I Got Pretty Close

PostPosted: Tue Jan 23, 2018 8:41 pm
by Mindfront
Thank you for taking time to do this detailed inspiring guide.

Making own vertex group assignments is the most tedious part in the entire clothing creation process, but necessary if the clothes have more details. When I did the ninja dress I spent most of the time trying different solutions and ended up with a total of 19 vertex groups distributed on left, mid, right and rigid, although I had quite a fair amount of experience at that time. In some cases I use both the helper and the body depending on which gives the closest desirable result, an example viewtopic.php?f=20&t=14345

MargaretToigo wrote:Once I have a good fit -- which I do with this suit -- I can fine tune the weight painting in Blender so the clothes will deform correctly when the human is posed.
Have you tried Blender's new Surface Deform modifier? A very good and fast help to make the clothes to follow the body without weight painting.

MargaretToigo wrote:I try to do semi-real for the clothes I share here and I usually provide spec and/or bump maps to help add to that semi-realism, but it's up to those who wish to make them even more realistic to do so. All my stuff here is CC0 so anybody who wants to can improve upon my work and share it far and wide.
It's very simple to make your "semi-real" clothes to look almost real if needed and I am very happy for your contributions as it makes the freedom of choice greater.

MargaretToigo wrote:
punkduck wrote:And a high resolution, like you used for the buttons, even creates worse effects than a low resolution.

Now, that is interesting as I would have thought it was the opposite, that higher resolution meshes would hold their shape better than those with lower resolution. But perhaps the more vertices there are in a mesh, the more vertices there are that can potentially be assigned incorrectly by MakeClothes.
I agree, in most cases the less topology the less weird deformations as so far experienced but then one have to think about the texture as when a low resolution mesh is subdivided the texture along the uv seams can be ugly deformed.

Re: MakeClothes Universal Fit -- How I Got Pretty Close

PostPosted: Wed Jan 24, 2018 12:49 pm
by MargaretToigo
wolgade wrote:
MargaretToigo wrote:Vertices assigned to *groups with three points of reference do behave differently from those that have only one point of reference,

All vertices of a proxy have three points of reference on the human mesh. In other words: Every proxy vertex gets influnced by a triangle on the human mesh. The difference between the regular fitting and rigid being that regular proxy vertices get individual triangles, rigid proxy vertices share the same triangle.


I saw this explanation earlier in the thread, and I remember reading something about it in the past -- probably somewhere on this very forum -- but it took until this most recent project for it to actually make some sense to me.

I kept seeing how there was something different going on when I used *groups and three vertices, and when I used regular groups and one vertex, but I didn't really understand what was going on, only that the results were different.

Now I can see why it worked so well to assign all the shirt buttons and the placket into one * group, with three mid point vertices on the chest of the tights helper. And why it worked better to put some jacket fabric vertices into the *groups for the jacket buttons than when I just assigned the buttons alone.

Experience is the best teacher.

Re: MakeClothes Universal Fit -- How I Got Pretty Close

PostPosted: Wed Jan 24, 2018 3:30 pm
by MargaretToigo
Mindfront wrote:Thank you for taking time to do this detailed inspiring guide.

Making own vertex group assignments is the most tedious part in the entire clothing creation process, but necessary if the clothes have more details.


I'm glad to be of some assistance. There isn't a whole lot of documentation on MakeClothes, especially how to use own vertex groups and *rigid vertex groups, so I hope this helps to clear up some of the mystery.

Asymmetrical clothing meshes with details and embellishments (like buttons and brooches) are all unique in structure and design, and each needs to be analyzed and tested individually.
The more experience one gets making various types of assets, the more effectively one should be able to analyze and test different vertex group assignments.

Mindfront wrote:Have you tried Blender's new Surface Deform modifier? A very good and fast help to make the clothes to follow the body without weight painting.


I've got Blender 2.79 installed, but I haven't played around with it very much thus far -- I've been rather obsessed with MakeClothes lately.
A lot of the new features seem to be about improvements to Cycles, which I don't use very often, but I will definitely have to have a look at the new Surface Deform modifier.

Mindfront wrote:It's very simple to make your "semi-real" clothes to look almost real if needed and I am very happy for your contributions as it makes the freedom of choice greater.


Well, you've got the photorealism thing down, the renders on your thread about skin textures are incredible. Getting the skin right is a large part of staying out of that uncanny valley.

MakeClothes forces you to create good topology. It won't even run if there are tris, n-gons or vertices with too many poles.
And even when you've got an all-quad mesh that follows all the rules, it's not gonna fit or deform well if its structure is noisy and/or inconsistent.

Mindfront wrote:in most cases the less topology the less weird deformations as so far experienced but then one have to think about the texture as when a low resolution mesh is subdivided the texture along the uv seams can be ugly deformed.


It seems that, like a lot of stuff in CG, it's a six of one, half dozen of another situation that depends upon what something will be used for and how much of it will be seen up close.

Re: MakeClothes Universal Fit -- How I Got Pretty Close

PostPosted: Wed Jan 24, 2018 9:34 pm
by punkduck
MargaretToigo wrote:Now, that is interesting as I would have thought it was the opposite, that higher resolution meshes would hold their shape better than those with lower resolution.
But perhaps the more vertices there are in a mesh, the more vertices there are that can potentially be assigned incorrectly by MakeClothes.


Here is an example of distortion using o4saken harley top1. On the left "average female", on the right my "Dani" character. The buttons were not created on rigid groups.

deform.png


MargaretToigo wrote:Everyone needs at least one formal conservative suit, regardless of their profession.


Hmm, I not sure. Funny enough, but I did never need one for my profession.
I guess my last one was eaten by moths ... and after that the poor animals die of starvation :lol:

But for my characters I really like to put them in different roles (like an actor), Evi was even train driver (or is that called "engineer" in the States?).
So why not wearing a formal suit instead of a bikini ... we have winter over here ;)