Author Topic: Real-world scale  (Read 294 times)

Is there a simple way to increase the real-world scale of an entire substance WITHOUT causing repetition in the tiling?

For example, Beech Toast Wood from Substance Source looks to be about 1mx1m.  How can I easily increase the scale to, say, 3mx3m without just repeating the original output maps 3x?

That really isn't about substance, more about your control of it in the final rendering software. When you export the image, you just adjust the UV mapping there to reflect the shift. So in Blender, if I had a 3M plane and it was UV'd to fit the whole plane, it would be 3x3, conversely, if the plane was 1M, it would be 1x1m. To make it look 3m on a 1m plane, I would shrink the UVMAP poly to 33%.

I'm sorry if I wasn't clear.

Within Substance, I'd like to shrink the final output maps and then have the remaining "empty" space fill in procedurally with new unique pixels.  The method you described would simply repeat the original map three times in each direction, resulting in visual repetition.  I'm looking for a solution that would avoid new repetition in the tiling.

Here's a quick diagram that hopefully better illustrates the desired solution.  If the first map is the original, I'm trying to achieve the third map (not the second).  Does this make sense?

As an arch-viz artist, it seems to me one of the biggest strengths of procedural materials is the ability to generate "more" of a material when needed, without having to simply repeat it.  Filter Forge makes this very easy.  It would be great if the same was possible in Substance.

Thanks for your reply!

Yes! This is exactly what I'm trying to do but it's a real pain, a lot manual work if I want proper results.

it have always been a problem in Substance Designer since day one. You have to link lots of thing , basically re-program each level of each subgraph, each fx-map and even after that it often doesn't looks real anymore.  Yeah, Filter Forge is cool in that regard. But it's way less convenient in many other regards and pretty slow cpu bound.  I once tried to do slope blur there and it took forever to calculate.

   Imo easier to just make graph that assembles few randomly generated pieces into a bigger one.   Also a pain actually since SD doesn't have same easy "2d transform"gizmo normal image editors have.

I asked for a usable transform2d gizmo for years , then just gave up

Last Edit: May 10, 2019, 11:33:07 pm

Yeah I tried to blend together 3x3 materials with different random seeds but the results are "barely alright-ish" at best. Even with fairly homogeneous materials like concrete the transition between cells loses contrast. And if you have macro structures like pebbles or wood grain then it's just a mess.
At a user level it's either hacking a tiling material to look less bad or start from scratch and do a material with the desired scope. A proper solution, a flexible, easy to use control over coverage would need lower level support in the software. And if it can't be done because of technical limitations then it would be nice to at least have Substance Source contain materials with a bigger scopes, not just these teeny tiny 1x1, 2x2m ones.

I totally agree, those Source material are of little help to me same as Megascan ones  although.  For the environment we are doing non square  textures usually covering  5x10 meters of  real world.  1x1 or 2x2  are repeating as hell and more interesting they look on the site shader balls  less useful they usually are.

To bad proper scaling is a nightmare in SD.  Same with scale factor.   If you need to do 2x1 or 4x1 scale you basically have to redo everything from scratch.  Even with "non square expansion" 

From little of what I have seen about Alchemist it doesn't look it's going to offer any help in that regard too 

 Surprisingly  Allegorithmic did so much to improve the soft, so many very useful nodes and minor improvements but somehow I guess because  very basic architectural limitations it still tremendously inconvenient  and tremendously artist unfriendly.   Hope Adobe would managed to do something with it.   
   No hope for Filter forge. It looks totally dead. Only minor cosmetic changes for many years.  Perhaps with 64 core threadrippers it would have a chance.  Somehow those new-coming  CPU could probably become more attractive than hi-end GPU I hardly see any performance increase in SD with.   

Last Edit: May 13, 2019, 03:14:31 pm

I still use Filter Forge for image processing but not for texture generation in the classic sense. But now that I think about it... if there was a quick way to generate an atlas of arbitrary size (2x2...10x10) where each cell is the same substance but with different random seed the that result could be fed to a FF filter which removes the seams between the cells... hmm...

If the thing is a roof tiles, cobblestones,  any repeating pattern by its nature SD could do it with flood fill and flood fill gradient  separating few  cobblestones  shared among all parts on the edges of atlas.
  But if it's something  from mother nature and irregular I use Artomatix .   Our firm bough a few seats.   IMO it's a future of texture making. An AI based tool.  Does a perfect seam and content aware fill  synced across normal/ height/color/roughness whatever.  Not real time but just a minute or so. Simple and painless.  Often I don't even have to regenerate new substance. It expands  or "mutates" it on its own, content aware style.
  Sometimes I trully wonder how it recognizes the subject and does the seam blending not just by blurring one thing into another but rather  "content-aware" too.  Even on height channel.
  Hope Adobe will do something like that someday
Last Edit: May 13, 2019, 06:46:31 pm

Ah, Artomatix looks interesting although the lack of pricing doesn't make it seem indie friendly...

It's pretty expensive.  And they don't sell to individuals unfortunately.  From what I asked last year at least.       

Then I guess I'll take a stab at FF if I can figure out a non-PITA way of generating random seed atlases in substance.

Since Artomatix is now offering indie licenses as well here is a quick article how to solve scope change using ArtEngine: