Page 851 of 853 FirstFirst ... 351751801841849850851852853 LastLast
Results 17,001 to 17,020 of 17043
  1. #17001
    Member Ace Dragon's Avatar
    Join Date
    Feb 2006
    Location
    Wichita Kansas (USA)
    Posts
    28,689
    Originally Posted by rawalanche View Post
    No it can not. I can not separate concave and convex areas, which is absolutely crucial. You want edge wearing to appear only on convex edges while you want dirt to only accumulate in concave areas. With this map, you can not either of these things cause it's not possible to separate the contribution. Let alone that it's slower than a proper AO map would be.
    Brecht did mention somewhere that a curvature shader is something he would like to have in Cycles, but there's also many other things he would like to have in Cycles.

    Unless we see high development activity from other devs. like Lukas Stockner, I can't say when a requested feature is likely to land.
    Sweet Dragon dreams, lovely Dragon kisses, gorgeous Dragon hugs. How sweet would life be to romp with Dragons, teasing you with their fire and you being in their games, perhaps they can even turn you into one as well.
    Adventures in Cycles; My official sketchbook



  2. #17002
    Member Felix_Kütt's Avatar
    Join Date
    Apr 2005
    Location
    Hiiu, Nõmme, Tallinn, Harjumaa, Estonia, EU
    Posts
    4,451
    Originally Posted by Ace Dragon View Post
    Brecht did mention somewhere that a curvature shader is something he would like to have in Cycles, but there's also many other things he would like to have in Cycles.
    If I understand what he wants correctly then a curvature shader alone is not really going to help him since it's only a part of what he wants - what would be missing from it would be derivative of proximity to other objects geometry. And if I misunderstand him then there's no issue since pointiness + colorramp should already offer him what he wants.

    @rawalanche: If you feel I still misunderstand you feel free to start a thread in the materials and textures section and PM me, lets see if we can't figure something out for you.
    FunLinks: . . . . . . .



  3. #17003
    Member SterlingRoth's Avatar
    Join Date
    Mar 2006
    Location
    Portland, OR
    Posts
    2,115
    Pointiness is a half solution at best. A per pixel effect is significantly more valuable than the per vertex effect we have now.



  4. #17004
    Member Felix_Kütt's Avatar
    Join Date
    Apr 2005
    Location
    Hiiu, Nõmme, Tallinn, Harjumaa, Estonia, EU
    Posts
    4,451
    Originally Posted by SterlingRoth View Post
    A per pixel effect is significantly more valuable than the per vertex effect we have now.
    Care to clarify?

    Per texel? Per screen sample, normalized to what unit space?
    FunLinks: . . . . . . .



  5. #17005
    Member SterlingRoth's Avatar
    Join Date
    Mar 2006
    Location
    Portland, OR
    Posts
    2,115
    Originally Posted by Felix_Kütt View Post
    Care to clarify?

    Per texel? Per screen sample, normalized to what unit space?
    Like the way the bevel node works now. The distance slider is normalized to world space. Though, an option to use local space could be handy in some circumstances.



  6. #17006
    Member Indy_logic's Avatar
    Join Date
    Nov 2008
    Location
    Portland, Oregon
    Posts
    1,018
    Originally Posted by rawalanche View Post
    No it can not. I can not separate concave and convex areas, which is absolutely crucial. You want edge wearing to appear only on convex edges while you want dirt to only accumulate in concave areas. With this map, you can not either of these things cause it's not possible to separate the contribution. Let alone that it's slower than a proper AO map would be.
    Wait what? Can't you just use it as a mask?



  7. #17007
    Member
    Join Date
    Jun 2010
    Location
    Melbourne
    Posts
    1,291
    Originally Posted by Indy_logic View Post
    Wait what? Can't you just use it as a mask?
    yes I but I think what he means is that the bevel mask takes concave and convex edges without distinction. He would like to have the ability to take only one of the two.
    Blender Cycles: Lighting and Rendering Cookbook - http://www.packtpub.com/blender-cycl...-cookbook/book
    ArtStation - https://www.artstation.com/artist/bernardo



  8. #17008
    Member
    Join Date
    Mar 2012
    Location
    Italy
    Posts
    2,070
    Originally Posted by brecht View Post
    I don't know at the moment. We've talked about it with other developers and all agree it would be good, but I don't want to make any predictions about when. For sure it's hard work to implement it as well as the SPI guys did.

    It's not so likely to help much with a typical interior renders though, results can be worse too when there are e.g. light fixtures in the way that make the light tree prediction fail. Making it work well with IES and portal lights is not obvious either, there's some extra challenges there not solved by this algorithm. Adaptive pixel sampling would probably be more effective to reduce noise in interior scenes, and personally I would give that higher priority.

    I'll add it to the GSoC ideas list in case any student is interested.
    Actually occluded lights are not so much in most of my scenes, the majority of them are rows of light fixtures on the ceilings of the interior which i solve by making the physical light fixture invisible to diffuse rays and instead using IES applied to point lights which loght everything.

    when doing luxury villas i easily reach tens of them, table lamps with occluded "bulbs" are the minority in my cases.. so these points lights with ies are straight visible to traced rays.

    But i can understand what you mean, not going to push for it, just asked. Thanks for clarifying!
    Testing: Vray for Blender // Old projects: #1 - Eames chaise lounge #2 - Chester Interior



  9. #17009
    Yes, as I said, the bevel is completely unusable for procedural shading as in the vast majority of cases, you need only masks of concave corners, or only mask of convex edges. Bevel map will give you both of these at once with no possibility of separating them.

    Pointiness is even less feasible solution, as it's dependent on underlying geometry. You can not generate edge wearing mask out of 6 quad cube for example. And adding additional topology onto my models just because of shading is not an option.



  10. #17010
    thanks Brecht for the new random walk sss.it gives very natural results.i have tested with the 2.79.2 Windows 64 build,and if i am not wrong,i think i found a bug.i have rendered with an lightly overlapping mesh and it gives black spots,the burly sss doesnt do that.
    just saying ,if it helps .

    here a pic.the mesh from the gum was mirrored and applyed.its slightly overlapping in the middle.

    sssbug1.JPG
    Last edited by pixelgrip; 14-Feb-18 at 11:44.



  11. #17011
    Member Ace Dragon's Avatar
    Join Date
    Feb 2006
    Location
    Wichita Kansas (USA)
    Posts
    28,689
    Pixelgrip; Part of the issue is that Randomwalk is far more dependent on the mesh being a volume (meaning that it is a bit less tolerant towards overlapping/non-manifold geometry). Brecht did mention in the log that it might be possible to mitigate to an extent, but not to the point where users would not have to make certain changes to their geometry.
    Sweet Dragon dreams, lovely Dragon kisses, gorgeous Dragon hugs. How sweet would life be to romp with Dragons, teasing you with their fire and you being in their games, perhaps they can even turn you into one as well.
    Adventures in Cycles; My official sketchbook



  12. #17012
    Member Felix_Kütt's Avatar
    Join Date
    Apr 2005
    Location
    Hiiu, Nõmme, Tallinn, Harjumaa, Estonia, EU
    Posts
    4,451
    Originally Posted by rawalanche View Post
    Pointiness is even less feasible solution, as it's dependent on underlying geometry. You can not generate edge wearing mask out of 6 quad cube for example. And adding additional topology onto my models just because of shading is not an option.
    Colorramp is a thing and so is subdivision modifier with the simple subdivision method.

    The bevel node can be separated into derivatives with separate XYZ node and re-combined with the combine XYZ node. modify the scalar derivatives as needed in between.
    FunLinks: . . . . . . .



  13. #17013
    Member SterlingRoth's Avatar
    Join Date
    Mar 2006
    Location
    Portland, OR
    Posts
    2,115
    Originally Posted by Felix_Kütt View Post
    Colorramp is a thing and so is subdivision modifier with the simple subdivision method.

    The bevel node can be separated into derivatives with separate XYZ node and re-combined with the combine XYZ node. modify the scalar derivatives as needed in between.
    We're well aware of the workarounds. They certainly do work, within their limitations, But they are not ideal. A proper AO node that allows masking, and a proper curvature node would be a great addition, and certainly an improvement over quadrupling polygon counts just to get an edge detail, which is still vulnerable to topological distortion:
    edges.jpg
    left:pointiness + subd (4 levels: 256x polygon count), center:pointiness alone, right: bevel edge hack.

    Note how a shader based solution follows the edges consistently, regardless of topology. Other than the lack of concave/convex separation, the bevel hack works well.
    Last edited by SterlingRoth; 14-Feb-18 at 16:07. Reason: smilies



  14. #17014
    Originally Posted by Ace Dragon View Post
    Pixelgrip; Part of the issue is that Randomwalk is far more dependent on the mesh being a volume (meaning that it is a bit less tolerant towards overlapping/non-manifold geometry). Brecht did mention in the log that it might be possible to mitigate to an extent, but not to the point where users would not have to make certain changes to their geometry.
    thanks for clarification,have not seen the log you are mentioned.its a known issue then.



  15. #17015
    Member Ace Dragon's Avatar
    Join Date
    Feb 2006
    Location
    Wichita Kansas (USA)
    Posts
    28,689
    Originally Posted by pixelgrip View Post
    thanks for clarification,have not seen the log you are mentioned.its a known issue then.
    It's less of an issue and more due to the nature of the shader itself.

    The scattering effect with Random Walk is done by using brute-force volumetric scattering inside of the mesh (similar to the Volume Scatter shader, but faster to render and more controllable). I mention the more controllable part because you can't control the density value in the Volume node with a surface texture for instance. You will find the same issue in Renderman and Arnold, two commercial engines that already have it.
    Sweet Dragon dreams, lovely Dragon kisses, gorgeous Dragon hugs. How sweet would life be to romp with Dragons, teasing you with their fire and you being in their games, perhaps they can even turn you into one as well.
    Adventures in Cycles; My official sketchbook



  16. #17016
    Member Felix_Kütt's Avatar
    Join Date
    Apr 2005
    Location
    Hiiu, Nõmme, Tallinn, Harjumaa, Estonia, EU
    Posts
    4,451
    Originally Posted by SterlingRoth View Post
    "workarounds"



    Originally Posted by SterlingRoth View Post
    which is still vulnerable to topological distortion:
    edges.jpg
    Pointiness + subd
    Originally Posted by Felix_Kütt View Post
    Colorramp is a thing and so is subdivision modifier with the simple subdivision method.
    And yes, I am expecting the user to have good topology. That is not unreasonable.

    Originally Posted by SterlingRoth View Post
    Note how a shader based solution follows the edges consistently, regardless of topology. Other than the lack of concave/convex separation, the bevel hack works well.
    Never said the bevel shader shouldn't be used, rather that you can use the pointiness to get a gradient to differentiate between concave/convex geometry. Nor am I saying I'm against more ways of achieving the results, the more the merrier. I'm just saying things can be achieved now where some here are claiming it's impossible.
    FunLinks: . . . . . . .



  17. #17017
    Member Ace Dragon's Avatar
    Join Date
    Feb 2006
    Location
    Wichita Kansas (USA)
    Posts
    28,689
    However, having good topology for the pointiness parameter might mean you don't have topology that is ideal for texturing and animation.

    In addition, if you're using micropolygon displacement or a heavy level of subsurf, Pointiness becomes useless (you would get very narrow lines at best). You can make it work with subsurf in some cases, but it can be a bit of a pain to tweak the mask whenever you change the subdivision level.

    The artists who did Cosmos Laundromat had an example of it being used on rocks, and for that purpose it can work rather well (but not in other cases like if you need wear across the geometry of an entire mech).

    It's great when it works though because it doesn't slow down rendering at all, but sometimes the slower methods that involve tracing inside of the object is the only real solution if you don't want to bake the wear or dirt into a one-off texture.
    Sweet Dragon dreams, lovely Dragon kisses, gorgeous Dragon hugs. How sweet would life be to romp with Dragons, teasing you with their fire and you being in their games, perhaps they can even turn you into one as well.
    Adventures in Cycles; My official sketchbook



  18. #17018
    Member Felix_Kütt's Avatar
    Join Date
    Apr 2005
    Location
    Hiiu, Nõmme, Tallinn, Harjumaa, Estonia, EU
    Posts
    4,451
    Originally Posted by Ace Dragon View Post
    However, having good topology for the pointiness parameter might mean you don't have topology that is ideal for texturing and animation.
    I assure you that would be very good topology for texturing, and it wouldn't be an issue for animating hard surface type object, nor can I see much reason you would want to utilize this shader technique on an organic model.
    FunLinks: . . . . . . .



  19. #17019
    Originally Posted by Felix_Kütt View Post
    I assure you that would be very good topology for texturing, and it wouldn't be an issue for animating hard surface type object, nor can I see much reason you would want to utilize this shader technique on an organic model.
    Pointiness is simply not an option, that's it. Let's say you have a hard surface model of about 2 million polygons. The model has large flat areas, so subdividing it evenly would make it let's say 30 mil. polygons. That will bring viewport down to a crawl, and we are talking about just a single one, now imagine a few of them in the scene. Even if I deliberately placed the topology only in the specific areas to make pointiness work, it would add a lot more topology, and would eat a ridiculous amount of time.

    The whole point of this procedural shading methodology is to reduce work time by reducing need for correct topology and UV mapping.

    Here's a wireframe screenshot.
    https://80lv-cdn.akamaized.net/80.lv...Wire_Front.jpg
    Do you think it's even remotely feasible to evenly subdivide something like this just because of materials?

    You can clearly see there are many areas where pointiness would simply not work. And the main point here is that the solution MUST be topology independent,so that it can be applied to any new mesh without any prior changes to it.

    Look, I am not looking for a workaround, because I am simply 100% certain there isn't a feasible one in Blender at this moment. I am just requesting a feature that should be trivial to implement and would enable Blender users to use incredibly powerful new shading techniques. I can't understand why anyone would be against that.

    The argument that you can use pointiness to differentiate between concave and convex is flawed for multiple reasons:

    1, Calculating pointiness and bevel shader in the same map network will make the already slow shader even slower.

    2, It won't work on for example simple 6 quad cube without adding additional topology.

    3, It doesn't reduce workflow time, on the contrary actually, it increases it.

    EDIT:
    I've made a video showcasing how AO based procedural materials work in practice:
    https://youtu.be/qIIaQmfrmsc
    You just model away whatever you want, and as long as there is some box texture projection applied on top, that's all you need. It just works, in realtime.
    Last edited by rawalanche; 15-Feb-18 at 06:22.



  20. #17020
    Originally Posted by Felix_Kütt View Post
    And yes, I am expecting the user to have good topology. That is not unreasonable.
    Well, a very large number of jobs these days consist of visualizing some sort of imported CAD geometry. This geometry is usually so complex that it is impossible to retopologize it in an even remotly cost effective way so you have to work with "bad" toplogy.
    Here is an example (imagine the following object and about 5000 similarly bad ones in one scene):

    A pointiness map that ignores topology would be immensly useful for jobs like that. Something like VRays dirtmap.
    https://docs.chaosgroup.com/display/...p+%7C+VRayDirt




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •