Colvillesque

Works in progress and technical screen shots.

Moderator: joepal

Re: Colvillesque

Postby blindsaypatten » Tue Jul 04, 2017 6:33 pm

Perhaps I have missed a specification, but for the cloth presets to have any meaning they really need to have some reference to the scale and mesh resolution, for example the silk preset values might be for a scale of 1 BU = 1cm with a 10x10 grid per BU. It doesn't make sense not to give units because a 1cm square piece of denim will behave differently than a 1m square piece of denim when dropped on the floor.

| think I now understand the interesting mesh pattern used for Nate's leather jacket. Since bending occurs along edges the simulator will produce sharp bends parallel to the edges of a square grid but larger bends diagonal to the grid lines. I suspect that the mesh pattern in the jacket produces a more anisotrophic bending than a rectangular grid does.
blindsaypatten
 
Posts: 586
Joined: Tue Mar 14, 2017 11:16 pm

Re: Colvillesque

Postby punkduck » Tue Jul 04, 2017 10:39 pm

Okay so applied subsurface subdivision is not needed before you do the animation. That will save time. When doing the render it normally makes sense.

blindsaypatten wrote:
On the other hand, so far the jeans have always exploded if I try to animate movement, I think that bending at the hip produces more stretch than I've allowed for. In earlier experiments I found that too much stretch (structural parameter) and the clothes either sag or actually ooze off the body like slime!


I had similar experiences with the cocktail dress. The dress jumps up and down on the body like it wanted to explode in a few seconds. I also tried, if that dress will slide off the character when I "open the zipper". When I pushed it a bit over the shoulder it works. But it starts with a jump before gravity gets involved ... it is one of the few clothes I did, which is working with cloth simulation.

Here I used cloth simulation to put it on a coat-hanger: http://www.makehumancommunity.org/forum/download/file.php?id=3651&mode=view

blindsaypatten wrote:Simulations are scale sensitive, I think due to the distance parameters in the Cloth Collision properties. I will have to try dividing those by ten, up to now I've been scaling my model up by ten. I usually export at 1 BU = 1m and with the default parameters clothes either hover away from the body or downright explode.

Once any part of the clothes penetrates the body or non-simulated clothing the simulation is almost certain to fail. The challenge here is poses, or intermediate positions, where body parts overlap, whether it be arms at the side so that the arm and the clothing penetrates the body, joints that have formed a crease when bent, or a finger momentarily moving through another body part/the simulated cloth.


The funny thing is, that when the grid of the collision object is not fine enough, the cloth could even disappear in the character. I've done this necklace

http://www.makehumancommunity.org/forum/download/file.php?id=4595&mode=view

with cloth simulation (nearly rigid) so the pearls do not stretch. But since one pearl is connected to the next one with only one vertex it works. So I wanted to see if the necklace follows the form of the body and put it 20 cm over the shoulder a let it fall down. The faces of the breast are rather big and cloth simulation uses the distance to the nearest vertices. Therefore the simulation works until the necklace found enough space between the vertices and then disappeared in the character. So I enlarged the collision distance, but now the necklace hovered above the skin. So in the end I applied one image of the animation and fixed the rest manually.

blindsaypatten wrote:I think the cloth simulation has enough features to deal with your zippers and so forth, but I haven't tried those feature yet.


As long as you connect them to the piece of cloth, then it will work. But in normal case it is not really necessary, you can save a few edge-loops when you have to do sharp edges, when the geometry only intersects instead of being connected. I had an example in the clothes tutorial, which does not work: http://www.makehumancommunity.org/forum/download/file.php?id=4135&mode=view. For zipper, buttons etc it is possible to connect it with only one vertex (the necklace is done like that). But in the assets I didn't do that, so when you start the animation, for the buttons, zipper etc. gravity will start to work.

But thanks for your comments. My reason for asking for your experience: I thought of historic clothes for Native Americans. They always have these wing-like sleeves. But without cloth simulation these clothes will not really work.
User avatar
punkduck
 
Posts: 728
Joined: Mon Oct 17, 2016 7:24 pm
Location: Nuremberg, Germany

Re: Colvillesque

Postby blindsaypatten » Wed Jul 05, 2017 1:59 am

punkduck wrote:The funny thing is, that when the grid of the collision object is not fine enough, the cloth could even disappear in the character. I've done this necklace...


I don't think that's the real issue, I think the problem occurs when the step size is too large and the cloth penetrates the object between steps. To check this I created a simulation with a floor consisting of a single large plane with a smaller plane that falls in the middle of the large plane, so no nearby vertices on the floor plane, and it doesn't penetrate. https://www.youtube.com/watch?v=8eTICYrG6r8

I think if you crank up the quality/steps per frame, or damp down the velocity the necklace will behave itself. I've had entire shirts go right through a body when the movement per frame was too large.
blindsaypatten
 
Posts: 586
Joined: Tue Mar 14, 2017 11:16 pm

Re: Colvillesque

Postby blindsaypatten » Wed Jul 05, 2017 6:18 pm

Because the former table was too high contrast, and thus distracting, I replace the material with a light wood texture. But I don't understand why the sides of the table legs are rendered so much darker than the front but the side of the table top is only slightly darker than the front. The table top and table legs all use the same material.
ApoScene7UHD.jpg


This is also seen in the material view of the 3D viewer:
Table3DView.png


Any suggestions?
blindsaypatten
 
Posts: 586
Joined: Tue Mar 14, 2017 11:16 pm

Re: Colvillesque

Postby punkduck » Wed Jul 05, 2017 9:18 pm

blindsaypatten wrote:But I don't understand why the sides of the table legs are rendered so much darker than the front but the side of the table ...

Any suggestions?


The legs are a bit dark, yes. Did you use a diffuse-shader only, or a glossy-diffuse combination, a fresnel? Displacement? Hard to answer ...

The front side of the top is very bright, so it seems, that the direction of the texture influences the appearance.
User avatar
punkduck
 
Posts: 728
Joined: Mon Oct 17, 2016 7:24 pm
Location: Nuremberg, Germany

Re: Colvillesque

Postby blindsaypatten » Mon Jul 17, 2017 1:39 am

Departing a little from the Colville style by using direct lighting and depth of field, I wanted to see what I could do with pose and expression. The character on the right came out alright, the character on the left I'm much less satisfied with. I also put a little more work into composition, departing to a greater extent from the scene I started with.
ApoStairs5HD.jpg

I then tried adding this cue to make the "story" in the scene a little clearer:
ApoStairs5AltHD.jpg

I need to work on the changes in cheeks and around the eye when a person is smiling, I hit the limit of what I can do with just the rig.

Hmm. I see that somewhere along the way I lost the lighting to make this look like a late night scene.
blindsaypatten
 
Posts: 586
Joined: Tue Mar 14, 2017 11:16 pm

Previous

Return to WIP (Work In Progress)

Who is online

Users browsing this forum: No registered users and 1 guest