Only I did give it the needed geometry, the issues I pointed out is how the algorithm currently does not allow for even application even after subdividing it with a modifier and how you can suddenly get lackluster results when all you did was a apply a few cuts (in places that doesn’t even affect the amount of geometry around the edge).
With the current algorithm, you can mess up your shader if you decide to tweak the geometry or make it a higher resolution, there’s no artistic freedom because it forces you to model a specific way.
but we need to do better than copying the ‘dirty vertex colors’ math as I’ve shown above.
Right!
However, though a “dirty v colors” math, this one works on the higher level of a subsurf/multires modifier. The V-paint dirt still works on the base mesh. And, I have the feeling, it works on 32bit, not sure about the v-paint colors.
Its usable for certain scenarios but not as a real AO/curvature shader replacement. Tried it to dirty up/scratch elements of a spaceship cockpit, limits are the dependency on geometry density and the limitation of topology, meaning its not usable, rightfully, for non-connected sparse geo.
Don’t be silly, in many cases having a dirt or corrossive effect that changes when other objects come near is absolutely fake. This actually helps a lot, is faster than AO, allows procedural effects, allows to find sharp angles and such. If we end up having an AO texture it will need to have an option to not consider other objects… I doubt this will come soon.
It doesn’t even work with the crevices in the same object, if disconnected topology. As I said it is expected from the nature of the attribute but limits its usability. And what do you mean dirt collecting between objects is a ‘fake’?
That’s why he’s not saying that.
el_diablo is talking about static objects that are separate but they were fitted together. Imagine a dead Transformer lying against a stone wall in the desert, sand and dirt will collect in the crevices of his form, (like they would any object that’s not moving but is composed of several pieces) since the creature use to be able to transform into a vehicle, its shape is made of many difference objects.
No one want’s dirt to collect on intersection during animation, were talking about non organic objects that are complex and were assembled together should be able to collect dirt pre-animation.
I was playing around with this too, to figure out what it does and how it might work. It’s not AO neather it is true curvature calculated from the true geometry. Results are highly depending on the mesh subdivisions.
Looks like some kind of mesh based AO within one object only.
Well I was only replying to this part “And what do you mean dirt collecting between objects is a ‘fake’?” I didnt account for everything he said above it or any other text lol
It would be preferable if those who see this as ‘perfect’ already wouldn’t try to drown out those with legitimate critique on its usability and what could be done yet.
This thread, last I’ve seen, was designed to discuss the usability of the new attribute, yes?
It would also be preferable if those who don’t see it as perfect don’t flood the thread with complaints of ‘this isn’t perfect, it should be perfect instead’
I agree that limiting it to a per-vertex effect does make it less ideal than a true per-pixel effect. However, the current implementation is working exactly as intended and is generating a lot of good feedback from users. It’s a lot like the colored wireframes, it exists, it works, it isn’t perfect, but how long should we hold everything up in committee to wait for perfection?
It would be great if there were a fully sampled cavity/edge shader, users would love that. But this per-vertex solution works well within its limitations.
Should we remove buffered shadows from BI because raytraced shadows are so much better? No, because they have their place. I think that a computationally free solution to detecting edge curvature is a worthwhile tool to have. The shader that you are asking for isn’t a modification to this attribute, but a completely different feature.
Right in the center of that node system, there is an “add” node that only has one input. Why is that? Is the second entrance different than the first? Or is it an old useless node you forgot to delete?
Good to see I’m not alone, although I probably was one of the first to mention a criticism on this thread. And no, I’m definitely not hating on this new node. It has it’s uses and it’s fast.
Just some alternate approaches are needed to achieve (from an appearance perspective) similar if not same results, as different ways of doing it may work slower or outright have some failings in particular cases. And what I mean by edge angle or distance earlier isn’t really edges as in the mesh geometry sense, but detection of where surface transitions occur in topology. Making “hot/cold” mappings of where convexity or concavity occurs.
I could think of a few alternate designs for these. Perhaps a node where a user enters a min/max angle and distance. Transitions that meet that angle criteria are shaded at 1 and fall off to 0 at the set distance. And another approach would just use a distance from where surface transitions occur, and 1 to 0 value is determined by the angle of that transition. Quite different approaches, but used for similar effects. (And likely to be combined with math nodes for tailored results.) But these node concepts also parrallel pointiness in what they’d be used for, so perhaps grouped together in the future for user convenience. (Essentially a whole node family used for worn edges and dirty crevices would be useful.)
Although I’m very short on programming (just managing basics for stuff like drivers), I suspect some of the code needed for this stuff is already in Blender. Both Autosmooth (for mesh shading) and Freestyle appear to already have ways of detecting changes in surface topology. Not sure how efficient their methods are, but certainly something like what those things are doing could be leveraged for purposes of shading sometime in the future.
Correct me if I’m wrong here (I don’t quite understand all the technical aspects of it), but doesn’t this have something to do with the pointiness node…? I mean, this cavity mask stuff can be useful to combine textures and materials based on the changes in the topology right?
Inspired by this new pointiness, Antonis also decided to add the opposite: a positive or negative cavityeffect. Watch the video below to see how it works, or watch it with Antony’s live commentary at1:06:45 in the weekly video.
I never said the per-vertex part was the issue, the issue is the current algorithm makes is hard to maintain the shading effect you’re after if you do anything to it in editmode (as it will likely change the result in various spots).
This can be solved by basing the effect on a fixed distance from the corners instead, preserving the effect regardless of edits (unless it involves the removal of faces near the edge).
