SOLVED: Cavity or curvature input map?
Hi,
I read the instructions on the Quixel Wiki page and it says that the curvature/cavity map is the most important input. I understand but to my understanding cavity and curvature are different things.
Cavity is a grayscale map but curvature can be a red and green map showing the convex and concave surface.
None of the sample projects contain cavity / curvature so I couldn't check it out by mayself.
Which one do I have to use?
Thanks for your help.
Peter
I read the instructions on the Quixel Wiki page and it says that the curvature/cavity map is the most important input. I understand but to my understanding cavity and curvature are different things.
Cavity is a grayscale map but curvature can be a red and green map showing the convex and concave surface.
None of the sample projects contain cavity / curvature so I couldn't check it out by mayself.
Which one do I have to use?
Thanks for your help.
Peter
Replies
Sorry for the confusion! First off, DDO creates this map automatically, for instance from the Normal map. Unless you don't have a normal map, you will actually rarely need to plug in a cavity/curvature map. Here's how the cavity/curvature map expected by DDO can look:
There's a lot of confusion related to both these terms. Curvature can be (as you've noted) both an RGB map encoding convexity/concavity in separate channels, or a grayscale map encoding convex edges as white and concave edges as black like in the image above. This map is often (erroneously) called Cavity, but more accurately a Cavity map can also be a white map with concave edges encoded as black. Feeling more confused? You're welcome!
- Teddy
My problem was that I use a high poly model and the baked normal show almost no useful information. In this case DDO needs a proper curvature map because it can't convert the tangent space noral map to a proper cavity, right?
It's also not appropriate for tweaking it in Photoshop.
So I can bake the required map in xnormal but I wasn't sure which one should I bake.
Thanks anyway.
Now if DDO could make use of a curvature map as input it would be an amazing tool to work with high poly models.
I looked into it a while ago for a low poly mesh, went as far as replicating the exact same curvature map that DDO generates just to see what DDO uses exactly. It wouldn't give the same results no matter what. Not even when the imported Curvature map was the exact same as the one generated. I'm not sure what's going on with curvature maps and DDO.
As for high polys, this is what Quixel put together a while ago: http://quixel.se/dev/issue01
@bac9-flcl
Do you use a curvature map as curvature input? or you use it as a custom mask?
The curvature generated in Substance designer seemed to be all right.
So it's very disappointing because if I have to use a high poly mesh it's very difficult to make a proper curvature.
so that does mean that creating ao and cavity maps from ndo and then using them as input for ddo is completely unnecessary since ddo basically does the same internally?
looks like I did a lot of unnecessary steps with creating cavity and ao maps so far...
although it looks like ao maps created from ndo seem cleaner than those created from ddo. (rich as well as standard occlusion generate shadows on flat surfaces). any idea where that is coming from? if ndo can do it, surely ddo should be able to as well
But I didn't experience any problem with low poly models.
I've tried a lot of setting for baking curvature map in xNormal and none work with DDO. Really ATM, DDO is realy fine tuned to work with baked tangent normals only to create its curvature mask. I don't know why they added a curvature input in the first place as nothing input in there, will work as intended. You'll need to use a lot of workaround to get decent result.
So from now on , for high poly models, i'll just wait for the curvature map baker that is supposed to be featured in (hopefully) 1.9
I don't know, grayscale background monochrome mode curvature bakes from xNormal seemed to work as intended for me. Default xNormal curvature settings are completely unfit, yes, but it's easy to get the look recommended by the DDO wiki and once it's there, the map works alright in the dynamask editor. Combining concavity/convexity bake from Knald also seems like a solution.
What I'd like Teddy/Eric to clarify though is why they insist on recommending NOT to use baked curvature map input on lowpoly objects. If I remember correctly, several threads pointed time and again that DDO won't pick true geometric edges into curvature by itself (and it's not reasonable to expect it to when it's not performing any raytracing like xNormal does), only picking it where normal map contains the curvature. For objects with bezels baked into normal maps auto-curvature from normal might cover most edges, but for any object with no baked bezels on edges, no curvature will be picked and you'll have subpar results. Seems strange to recommend ignoring curvature input, especially when it's so easy to bake in xNormal or Knald.
I don't buy the argument of curvature interpretation in DDO being "fine tuned" for auto-generated input. Curvature is a middle gray layer with darkening on concave and brightening on convex edges, there is no hidden complexity in that, no place for hidden intricacies like in tangent space normal maps. Okay, you can bias it to get white/black extremes faster or slower, and you can configure the bake to get different influence radius from edges, that's it. And even if those two factors aren't perfect in your bake, you can compensate them with slight adjustments of dynamask parameters. I don't see how baked curvature would manage to ruin anything, in other words.
Again, i was never able to get a curvature map baked in xnormal to work properly in DDO
Here's a typical result i get when i look at what DDO creates as curvature mask from a curvature map. I've put each mask from dynamask in vertical slice. Sharp, fine and soft works well but medium, big large and huge are useless IMO. DDO adds way too much contrast for the more blurry mask.
Now curvature mask made from baked normals will be perfectly shaded every time.
@bac9-flcl I'd be curiuous to look at one of your DDO file to see how the curvature mask looks like in dynamask.
This issue you're having with the big black and white blobs is what I keep getting even when importing a curvature map that is the exact same as the one DDO generates which makes no sense at all. That's why I think the issue with this might be lying somewhere else.
If anyone else could confirm that they are getting the same results as me then that'd be massively appreciated. This is the original thread where it's discussed how to generate a curvature map from the normal map in the exact same way as DDO: http://www.polycount.com/forum/showthread.php?t=148357
I just had this idea to use Teddy's example of a curvature map and use it as a curvature input with a random mesh and created a project. Now when i open dynamask and check what DDO did as curvature mask, i see this
So even with a perfect curvature map, you still get too high contrasted mask (huge,large,big and medium). To me the problem lies in the script DDO use to create curvature mask from the curvature input.
- Teddy
I currently working with DDO for VFX and HighPoly stuff.
Really interesting post and tips here, thank you guys.
When you work with HighPoly model you don't have Normal bake or any of this stuff.
Everything you need for generate nice mask it's the famous Curvature map.
Thanks to Martin7, I finnaly understood how DDO works with curvature and found solutions.
First, yes it's seem that the algorithm to interpret external curvature maps in DDO is a little bit too strong.
Knowing that, we can made ajustment on setting when creating Curvature map and also adjust the map on Photoshop before importing on DDO.
Well, anyway. I will make some test etc and share with you my stuff and reschearch. Like I aways do on other workflow for my Project Wayne.
Please let us know when you do something with ddo and post it in the gallery thread!
Are you planning to release this update with 1.9 or that you'll make it available as a small patch before? Would be great not to have to wait till 1.9 for this.
I dont know if it can be usefull, i thought it might help the developpers to refine the script for the curvature map input. As you can see the masks look similar and very workable on both side up to medium. But then the large and huge mask from a curvature map input, are getting way too contrasted to be of any use.
Hope this help.
Would you be able to post any examples of the baked curvature maps from the GPU baker any time soon? Particulary interested in seeing how curvature from organic models looks compared to other software.
Will the GPU baker work similar to Xnormal in the sence that it does not display the high poly, allowing you to bring in from zbrush meshs/objects with millions of polies and baking f.e. 50-100 objects?
My gripe with Xnormal is that we have to explode the meshs and this creates issues with shading for ao, bent normals, prtp maps.
I like how I can bake in 3ds max and keeping the model together and seperate each element of each object by material ID and in RTT check hit only matching mat id's and go. But theres no way to bake such hp obj's in max unless it is done one at a time and even then it is painful on a high end comp.
Also I would like to know if the gpu baker uses Anti-aliasing such as 1x-4x as xnormal does or even higher....
So.. I was curious about the GPU baker that is coming.Thank you.