Ok ... Can we fix either the 3d displace or make the 2d displacement support UDIM ?

[Bigguns]

Hi there,

The 3d displacementmod in Vray can't be used in production for character because the precision of it change based on the camera distance.. so the futher away, the less details the displacement has ... and it fuck's up all my reflection, since less micro breakdown, my character look like he's out of the shower when rendered far from the camera, look's ok at bust shot and look's dry as hell in close up... can it be fixed soon ? I'm asking for the 2ddisplacementmod to support UDIM for 4 years.. still not there..

The 2d method, wich is the best displacement method I've ever seen, precise, fast, low memory and CONSTANT , close or far away from the camera, the details are the same, so it don't fuck up my reflection like the 3d mode... even max displace don't fuck up at distance ( but can't be use since too memory heavy ) .

Can you Vlado , please, put these issues on fast track and get it fixed once and for all soon ? Everyone I know using UDIM have the same problem... they up using only one 8K for their character to use the 2d displacement since the 3d is unusable ... can't make a look dev with it, it always change the result ofn the spec! It may work on simple details, but for small crips hi frequency details, the 3d displace fuck's it all!

Now Vray has great skin with Alsurface, will have good hairs with Alhair, can we please have a working 2d displacement with UDIM and fixing the bug on the 3d displace thing ? I tried everything , view dependant checked, unchecked does nothing etc.

And for the Subdivison method, wich I never used.. I tried, on my side it does absolutely nothing.. zero displacement coming out of it.. switching from 3d to subdivision = no displace.. is there a magic trick to make it work ? I'm using a 32 bits EXR .

I joined an image to show you, the face rendered at a bust shot distance, and one maybe twice more far away... and already you can clearly see the difference.. if I do it more away from the camera, she get's really like she's sweating as hell .. image that in an animation... would look retard as hell lol.

Thank you ! I'm sepaking in the name of many, trust me ^^

[^Lele^]

Hi there,

The 3d displacementmod in Vray can't be used in production for character because the precision of it change based on the camera distance.. so the futher away, the less details the displacement has ... and it fuck's up all my reflection

You do realise you can set the displacement modifier to a number of world-space based modes?
F.e., if you set edge length in world units in the 3d displacement mod, or if you used a world edge length in the subdivision modifier (with your displacement texture in its slot, as usual).

Contrary to 2d displacement, those modes will also be able to create static geo for you, making rendering a LOT quicker (provided you have the ram.).

[vlado]

Hey Luc,

As we discussed on Skype, it would make the most sense to get the 2D displacement to work with UDIM textures. I'll get to that once I'm back in the office and things settle down a bit.

Best regards,
Vlado

[Bigguns]

You do realise you can set the displacement modifier to a number of world-space based modes?
F.e., if you set edge length in world units in the 3d displacement mod, or if you used a world edge length in the subdivision modifier (with your displacement texture in its slot, as usual).

Contrary to 2d displacement, those modes will also be able to create static geo for you, making rendering a LOT quicker (provided you have the ram.).

F.e., if you set edge length in world units in the 3d displacement mod : I just can't see anywhere an option where I can change it World ... and if you meant by checking view dependant on and off, it does nothing to correct my problem. Where do you see that option world ?? I don't see it anywhere.. I'm using vray 3,4 right now. EDIT : So you mean checking the Bbox ?

Vlado ! Yeah awesome !!!! All charater modeler will be very happy to know that trust me !!

[Morbid Angel]

F.e., if you set edge length in world units in the 3d displacement mod : I just can't see anywhere an option where I can change it World ... and if you meant by checking view dependant on and off, it does nothing to correct my problem. Where do you see that option world ?? I don't see it anywhere.. I'm using vray 3,4 right now.

Vlado ! Yeah awesome !!!! All charater modeler will be very happy to know that trust me !!

First of all, we get your point without using strong language, its not necessary

For the view dependent vs world it works in a way where if you uncheck the view dependent, then it automatically becomes in world (system) units. In this case, you will preserve all the detail you are after the edge length become = unit length. To get fine detail your edge length will have be less or equal that of the smallest detail on your sculpted model. However if you are using tiled exrs, you might also lose some details, so using untiled un mipmapped textures would be the most accurate (but longest) way.

[Bigguns]

First of all, we get your point without using strong language, its not necessary

For the view dependent vs world it works in a way where if you uncheck the view dependent, then it automatically becomes in world (system) units. In this case, you will preserve all the detail you are after the edge length become = unit length. To get fine detail your edge length will have be less or equal that of the smallest detail on your sculpted model. However if you are using tiled exrs, you might also lose some details, so using untiled un mipmapped textures would be the most accurate (but longest) way.

ho ok, sorry if it sounds harsh, was not intented

So far, checking the relative to Bbos, is what helped the most. view dependant did nothing to keep sharp details at distance at all in my test, but Bbox helped though, but, still not exactly as detailed as in big, while the 2d displace is perfect in that regard.

What do you mean for the tiled EXR probably I don't get you'r point correctly , why would I loose details compare to one map ? I used four 8K maps for the bust, it's way more resolution then one 8k map and it's EXR, no compression so no losss of details. Can you explain more what you propose please ?

[^Lele^]

What Dmitry explained.
Further to all that: any method which takes you to the finished model from the base surface which is not (algorithmically) IDENTICAL to the one used at the Asset creation time (ie. non-adaptive CK subdivs, quad splitting, whatever) will by definition incur in a form or another of interpolation: even contrast sensitive, adaptive schemes will risk missing detail, or generate it where it didn't exist, and not where it did.
The *only* technically correct way to get your displacements exact is to match what base topology + subdiv scheme you built at sculpting time, so to have the displacement map push and pull on exactly the same vertices (and ideally vertex number!) as your zBrush, or Mudbox, or what have you.

That the differences may be invisible with enough (view-based at that!) detail, doesn't make it the right, or most efficient approach, nor would UDIM support for the heightfield mode (which i *really* would not use for characters, but hey, to each its own.).
That's why the displacement modifier can be used like a classic CK subdivision, with a low edge length and (so to force v-ray to subdivide all the way to a) specified number of recursions, for example.
That's also how most other engines tackle it (CK), whether in screen or world space.

[Bigguns]

So this :

The *only* technically correct way to get your displacements exact is to match what base topology + subdiv scheme you built at sculpting time, so to have the displacement map push and pull on exactly the same vertices (and ideally vertex number!) as your zBrush, or Mudbox, or what have you.

If I understand clearly what you mean, I guess I'm ok on that, because the base mesh I sculpt my skin pore details in Zbrush, is the exact same as in max, and in Zbrush , before sculpting I do store the base mesh as the morph target, so when I generate the displacement map, I do a switch morph target to generate the map from my base mesh, wich is identical as in max. I always did that, and work's well. I got exactly the same thing as in Zbrush... except when Zbrush have a hard time on twisted surface, like the ears.. it's why, since a long time I model the ears completely in max, just to do very minor tweak in Zbrush later on that I know would not mess my ears.

I don't know why you say you would not use the 2d displace for character, but it always gives me awesome result, the details are super sharp, crisp, and consistent, wich we wanted, it create nice breakdown in the reflection, way crisper then the 3d mode. And the 3d mode, when the model is far away, the displace loose resolution so it makes the model look wet, but now, with the Bbox check, it's way better then it was, but still not perfect, but meanwhile the 2d displace is supporting UDIM, I can live with that.

[Morbid Angel]

the tiled exr, basically segments the map into smaller lods, and choses automatically when to load those lods based on camera distance. So, it loads smaller versions of the same map which means the sharp detail is resized down, and further you get away from the model the more detail is lost.

Unmipmapped and untilied texture, will remain the same at any distance.

[Morbid Angel]

I found this image, its old but kind of explains the point.

So basically once you mipmap any texture, it holds those resolutions inside of it self. As the camera moves away it loads smaller versions. Its efficient because you have less IO over network, so faster load times. Its also less in ram, and chunks can be loaded/unloaded much like vray proxies. When dealing with hundreds or thousands of textures, this is the only way to render assets.

[^Lele^]

If I understand clearly what you mean, I guess I'm ok on that, because the base mesh I sculpt my skin pore details in Zbrush, is the exact same as in max, and in Zbrush , before sculpting I do store the base mesh as the morph target, so when I generate the displacement map, I do a switch morph target to generate the map from my base mesh, wich is identical as in max. I always did that, and work's well.

I don't know why you say you would not use the 2d displace for character, but it always gives me awesome result, the details are super sharp, crisp, and consistent, wich we wanted, it create nice breakdown in the reflection, way crisper then the 3d mode. And the 3d mode, when the model is far away, the displace loose resolution so it makes the model look wet, but now, with the Bbox check, it's way better then it was, but still not perfect, but meanwhile the 2d displace is supporting UDIM, I can live with that.

Base mesh plus displacement map isn't the issue: the issue is how you go at rendertime from your base mesh, to the displaced mesh topology.
Bad approaches will just crank meshing up, and sure enough, with enough detail, there is likely going to be a vertex just in the place where the original one was.
That's what your 2d does: it's a complete re-abstraction of the base mesh, and by virtue of its approach is able to generate all the detail you may want, and more.
Same can be said for view-based 3d displacement, when kicked hard enough for edge length.
Use a CK subdiv scheme while sculpting, and use the same CK settings with the vray displacement modifier at rendertime, and you will get EQUIVALENCE between source and destination topologies.
If 2d was giving you the sharp details you crave, but couldn't achieve at sculpting time, you are sure to have an issue to fix in workflow (ie. those are artefacts, not detail, unless your dispmap has more detail than the sculpted mesh...).

[Bigguns]

ha ok, I get you'r point now when you say that :

Unmipmapped and untilied texture, will remain the same at any distance.

But I'm not using any mipmapping , I'm just using a cmopsited texture witht he four 8K map inside of it, so it should be ok no ? Same set up as if I use one displacement texture, but instead I have 4 maps.. and it does the same if I use one udim instead of the four BTW.

[Bigguns]

Base mesh plus displacement map isn't the issue: the issue is how you go at rendertime from your base mesh, to the displaced mesh topology.
Bad approaches will just crank meshing up, and sure enough, with enough detail, there is likely going to be a vertex just in the place where the original one was.
That's what your 2d does: it's a complete re-abstraction of the base mesh, and by virtue of its approach is able to generate all the detail you may want, and more.
Same can be said for view-based 3d displacement, when kicked hard enough for edge length.
Use a CK subdiv scheme while sculpting, and use the same CK settings with the vray displacement modifier at rendertime, and you will get EQUIVALENCE between source and destination topologies.
If 2d was giving you the sharp details you crave, but couldn't achieve at sculpting time, you are sure to have an issue to fix in workflow (ie. those are artefacts, not detail, unless your dispmap has more detail than the sculpted mesh...).

Honestly with the 2d displacement, I dont get any artefacts, I did many many close up test render in the years and always found it really precise and crips. no artefact etc. As for the vraydisplacement in subdivision mode.. it just odn,t work here.. certainly my falut since I never used that mode.. do you have a scene to send that work so I can test it ?

[jstrob]

Base mesh plus displacement map isn't the issue: the issue is how you go at rendertime from your base mesh, to the displaced mesh topology.
Bad approaches will just crank meshing up, and sure enough, with enough detail, there is likely going to be a vertex just in the place where the original one was.
That's what your 2d does: it's a complete re-abstraction of the base mesh, and by virtue of its approach is able to generate all the detail you may want, and more.
Same can be said for view-based 3d displacement, when kicked hard enough for edge length.
Use a CK subdiv scheme while sculpting, and use the same CK settings with the vray displacement modifier at rendertime, and you will get EQUIVALENCE between source and destination topologies.
If 2d was giving you the sharp details you crave, but couldn't achieve at sculpting time, you are sure to have an issue to fix in workflow (ie. those are artefacts, not detail, unless your dispmap has more detail than the sculpted mesh...).

Hi Lele, about 2D versus 3D I did that test that shows what Bigguns is talking about, He convinced me that 2D gives more details when we use 3K and more for the resolution. Though 2d is slower, it will not freeze the system like when we go to small in edge lenght with 3D. You don't get those problems?

And about your CK subdiv, I'm not sure what you mean (did you mean Classic Catmul-Clark). To be sure we use the same CK in Zbrush and max/V-Ray what should we do?

[^Lele^]

Hi Lele, about 2D versus 3D I did that test that shows what Bigguns is talking about, He convinced me that 2D gives more details when we use 3K and more for the resolution. Though 2d is slower, it will not freeze the system like when we go to small in edge lenght with 3D. You don't get those problems?

Not particularly, no.
I can (well, hell. could. ^^) well budget my displacement, both for detail amounts and placement, using the settings which help with those type of issues: max subdivisions to cap excessively low edge lengths, world-space subdivision for hero stuff, lower quality view-dependent 3d displacement for environments and such.

And about your CK subdiv, I'm not sure what you mean (did you mean Classic Catmul-Clark). To be sure we use the same CK in Zbrush and max/V-Ray what should we do?

Yes , I meant Classic Catmull-Clark, purely because it is a standard, and as such results can be inequivocally compared, and when needed, matched.
This said, i think Mudbox may be implementing a CK method, but am not sure ZB does anything quite standard in that department.
However, i am no expert, so you'll have to do your own digging.
Further to this, explicit 3d methods have the benefit of being deformation-invariant, when properly set up, while the 2d method is hardly fit for that purpose (not by chance "Landscape" is written by it.).