Still Displacement Problems. Rules?

[dinf]

Hi everybody,

I still encounter problems with Octane's Displacement. Mostly they don't appear with smooth geometry like spheres or even strong deformed objects with over 90° edges. But using a standard cube or extruded splines will always tear at the edges as long as ready textures are used. Indeed I want to texture a simple wall made by an extruded spline, and this wall is very simple, with no or probably one bevel-step at the edges. In both cases the edges tear appart.
I'm not sure to understand the rules in wich cases this works and in wich not since cubes with a bevel won't work, an over 90° deformed plane works. The planes will show displaced edges at 90° wich aren't on the map, but they don't tear appart. I tried with 8/16/32 Bit b/w maps as well as procedural C4D shader wich aren't my goal, seems to make no difference, surely less than the polygon's angles. And I could paint all edges in BP in same gray tone, but I don't want to, this would delete the effect.
So what is the best way to do displacement?

[shawnfrueh]

Displacement is still 'broken' in octane. What they have at this point is an incredible starting point. From what I can tell Octane is displacing the actual polygon and not the points. This is why it will look like the edges are separating. The separation happens at UV edges, I unwrap my objects accordingly such as to hide the seam as much as possible if the displacement is acting up. If you search the standalone forums you will see a lot of post pointing this out. It also helps that, if you can, level out your displacement map along the edges to black. Its tricky to get to work right but the fact is that what they have working at this point is still incredible and they are one of the first GPU renderers to have this support.

[dinf]

Hey Shawn, thanks a lot.

The separation happens at UV edges, I unwrap my objects accordingly such as to hide the seam as much as possible if the displacement is acting up.

I've read about 2h in the forum to get a glance at the rules, if any. But in my tests I can't see that the UV seams are the bad guys only because I get partially connected renderparts at the seams as well as partially broken parts at connected UVs. In both cases there are 90°, and in both cases there are UV seams, connected or not. Thats the reason why I'm asking for rules. Oh, and I wrote a false one. What I wanted to say "I'm not sure to understand the rules in wich cases this works and in wich not since cubes with a bevel won't work, an over 90° deformed plane would do." This confirms your statement, the broken edges at connected UVs mine ^^

So painting all seams and nonseams at ~90° seems to be the surest way to get a floating presentation of the displacement (that works), beneath hiding all seams.

[dinf]

Good start, right, it was. 1 and 3/4 year later and still the same issues with displacement, right?
Is there planned anything about it in near future?
I have a bunch of objects wich are needed in a project, way too much to work them all over in this or that way.

[Terryvfx]

Yeah the displacement in Octane it's far from perfect. I've yet to try this but maybe this could help you:
http://polygonspixelsandpaint.tumblr.co ... 9733351349

A technique like this would certainly be much more powerful if we could have support for more than one uv set

[dinf]

Just saw this thread:

viewtopic.php?f=24&t=55669

Seems to be on it's way. How far to 3.1? Should have been in summer if I remember right, but to me it is better to wait a bit more until everything works well.



Sorry terryvfx, forgot to say thanks for the link They talk about seams. Them are not my problem but 90° angles and more. The displacement tears appart at these edges, they doesn't have to be a seam. I hope very well that Otoy will fix that in near future or, maybe, with 3.1