Minienv/diorama : PBR and material definition practice

nizu

Update : current status :



Original post :

Hi everyone,
Long time since my last post, but always coming back for inspiration , amazing works around here !

I'm working on a personal study. Below : current wip status , lowpoly rendered in Marmoset with NM details (baked from sculpts or hipoly chunks).

Diffuse map is a temp base , using AO and cavity.

Everything done in Blender + gimp.

There's still glitches in bake , and parts that could be refined (at least in sculpt brushwork)

Then the main focus for this will be to practice texturing and good material definition with PBR shading.

Also practice texturing techniques similar to DDO or Substance automations .. meaning dirts , edgescratches and acid effects done 'procedurally' from ao/nm/cavity maps..




More soon!

Obscura

The rectangular window couldn't work. there is just a thin wall behind it, and the stairs so it doesn't make sense to me. Maybe you should use a bit more geo here and there, the big rock and the damage on the top of the building is really angular. Also the uv seams are really visible at a lof of places. I know its wip, but these are the things that I would fix first.

nizu

Oh wow , no .. that window makes no sense, i totally missed that ! facepalm.

Thx Obscura . that's why i posted , i just finished a similar quick project (will upload later) and by the end i wasn't satisfied just because focusing on technical learning and workflow .. i overlooked too many other things for it being good portfolio material.

Good point on some pieces being too angular .. i'm actually pleasantly surprised i could bake sculpts with so much difference to a very lowpoly mesh , but it's "out of proportion" (i mean too little detail on lowpoly).
I guess one has to go back to lowpoly after scullpt to match it better to the changes , especially when chipping stuff .. i mean it would take ages to properly 'retopo' for this kind of env stuff -i guess- but at least a pass of corrections is necessary..

nizu

Update.
Replaced the impossible window, refined the sculpts of rock and walls, and fixed most projection glitches in AO.

Lesson for next time .. i should have broken the main object (walls and stairs ) into more pieces (split walls from floors or other walls) to avoid interferences (meaning those glitches from hipoly parts getting projected to some other nearby surface besides the one that should get the projection)

So far, in baking hi to low poly the only real "trick" i found is to split the model into simple parts. Maybe i'm missing something, but i never use a cage , i don't see how that would help or make work faster (having to make a 3rd mesh for baking ?) ... So seems to me splitting in small objects is simple , but actually very reliable way to manage bakes.

Spoon

Great start I like it, even though it is simple, and you have opportunity to show some sexy materials here.
I kinda want to delete the big round rock in the foreground though. I dont think it adds to anything.
Why do you include it? Does it serve a specific purpose?

Dave Jr

Looks great! - I agree with spoon;
delete the rock have some kind of detail prop there; a statue, a signpost or something.. something of character

nizu

Spoon , Dave jr : Thx and good idea.. Despite the sculpt refinement that rock still looked like an oversized pebble , and out of place.

I've replaced (most of it) it with a small watchtower .. it's not that original or characteristic but it's a good test case ...last year i textured a stilt house model using ddo , and it actually worked great .. with all those wooden sticks painting or working on the unwrapped texture would be a pain , but with a good sculpt base and some procedural dirt it shouldn't be too problematic.



This is the model now with fake-colors mat-id map . mostly baked quickly from materials assigned to hipoly chunks , but for some parts like the rock 'insets' in the lower left.. i had to hand paint it to match the sculpt.
Again this thing (mat id map) i learnt from working with DDO , but i think it's a good workflow in any case : 1st having a nice and complete mat id map + tangent space nm + AO + cavity)

martygalinskas3d

i agree with the big rock in the front should be removed/replaced. also the shingles should be hollow at the end and you could add a gutter with a little drainage duct. the wood texture on the boards seems a little large to me and where the wall is broken i think some broken boards would be nice as well. that's just my 2 cents.

Jack M.

I will try not to reiterate what others have already said, but I do have a couple of things to add.

- The rocks along the side of the wall could be pushed out more to give a better read of the silhouette. Because right now it kind of just looks like an oddly shaped wall rather than a wall being built on/next to some rocks.

- I'm also not a huge fan of the stand on top of the rock. I think it's blocking a lot of your hard work on the stairs and the wood panels in the back. It's doing somewhat of a disservice to your piece currently. I would say a statue with your best effort would do well there. That's where my eye is first drawn to anyway so putting your best piece there would be a good idea in my opinion.

nizu

martygalinskas3d,Jack M. : Good points , i've tweaked the lower rocks silouette , added thinner cracks to the wood , bumped inwards the shingles end.

As for the more significant changes you suggested .. i sort of set a time budget for this , if i have time later i could add a new texture sheet with detail props like small rocks , foliage, pipes etc.. and maybe a hero-prop : a high detail telescope for the watchtower.

About the watchtower blocking the view.. i am concerned about that , but otoh it's not hiding much (same as what's around actually ) and also a bit of depth and parallax should work nice especially in turnatable anim..


Base tileable textures :
As an experiment i baked this from a material in cycles , using the mat-id map to assign different tileable textures , mapped with a blended box map .. It seems worth the effort compared to just slapping tileables on the uv canvas in 2d .. no seams , easy to preview scale (and in case , it ignores/corrects distortion and difference in uv islands relative size..)

This isn't very representative of the final general colors unfortunately.. for the plaster yes, but for metal or most other painted surfaces.. the real main color will come later when i have dirt effects ready ..



Dirts test :

These are some dirt effects i'm preparing (thin dark dirt in cracks and creases, bright edges "spatter" effect, large and small "acid" or "paneling" effect)
I've made this starting from AO and cavity map, using Blender node compositor to reproduce (to the best of my knowledge) the kind of effects DDO does .

I have these dirts as b/w maps , this image is just a test using them as colors over the whole model .. i'll obviously then make different combinations for different materials ... but i need first to study more the effects themselves (how to detect more precisely concave or convex edges, how to make leaks,etc.. )

nizu

Update, 1st pass of diffuse using those bases and dirts (and the mat id map as 'selector' to change colors, contrast , alpha etc. to adapt each dirt to different materials )
At least this part of the workflow (which is pretty much standard ) went smooth and didn't give me much to think as other steps
I mean there are all sorts of variants , but the ingredients are always similar , in Gimp i assembled toghether layers (from tileable images directly or baked box mapped tileables ) and layer masks (dirt effects maps previously prepared in Blender node comp or Gimp itself) , then edited and tweaked anything necessary often using the mat-id as selector.
I've recently watched a Substance tut, and the general idea was pretty similar , though you can make full materials as 'substances' in Painter you can also use tileables as bases , then mask them with "b/w map" substances which are very similar to the "dirt effects" i'm using as masks.

Of course i'm aware of the other features of Substance (especially in a big complex pipeline , but before getting to think of that , the point of this test for me is to get a better understanding of the general ideas .. which are surprisingly independent from the app.

I mean : once you know the look and the 'math' (or process) for a good looking rock ... the matter becomes : what's faster or more reliable or gives better result using fewer or more applications , working more on a 2d canvas or direct 3d paint .. and how to speed up model detailing and baking ..
So many cool techniques and workflows out there , it's getting complicate :poly122: so one step at a time


Anyway, regarding the materials and texture result themselves.. i'm kind of happy with the plaster, the metal needs checking with a spec/gloss map and the wood and stone could use a re-work .. (prob. mostly in the base for rock , and in the dirts for wood..)

nizu

Update with full textures (spec and glossy)

The thing that took more work and re-doing is the wood .. with too much spec it just looks metallic , with little spec, it's boring .. So i tried on the panels on the back to have just some specks of shiny (glossy) , but -as for all other materials except metal- i had to tone down the "edge highlighting" (i do that with cavity map overlayed on everything + a layer of "spatter" bright specks along edges .. very similarly to DDO presets)

This edge highlighting looks pretty nice on almost any type of material in diffuse .. but when you go to spec and gloss, it can easily make anything look metallic..
Infact the plaster has quite a bunch of spec .. just not along the edges but in random specks along flat areas..

razz

This is progressing very nicely. I like the scene you're making. Stone walls seem to have quite bright edges and just as a preference everything a tad too monochromatic. Looking forward to see more!

WarrenM

I don't love the wood pieces stuck on the wall below the roof piece. It doesn't look right to me. Maybe make that a different sort of plaster or something?

Also, I think that some good weathering and grunge would make this piece pop out really well. Right now it's a little uniform in coloration and lighting.

Sebvhe

This is great work !

I just feel the rocks look a little bit off, like the fake concrete rock you see in amusement parks. It's a bit blobby.

But everything else is great

teaandcigarettes

As others have mentioned, I think you could greatly improve this by working on the sculpt a bit more. To get the extra sharpness, you can do that really quickly by looking at the mesh from the side, then using Zbrush's Clip curve to slice the mushy part off (you can do a straight line by holding shift if I'm not mistaken). It should be a relatively simple change that will only require you to bake your normals and AO again, but it will improve things a lot.

To give you an example, I've always been a big fan of Ilya Ndyal's sculpts. Notice the contrast between the clean areas and the damaged ones and how it helps make everything more readable.

Furthermore, I think you could drop the edge highlighting completely. With PBR, I it is better to rely on gloss and sharp edge normals to achieve desired result. Cavity maps can often ruin the look.

nizu

Thanks for the feedback everyone , this thread was on hold due to work but not forgotten ! I might turn it into a general sketchbook , but i'm trying to keep my tests and practice coherent and into few pieces, so i'll rework and expand the minienv.

Fro now , some quick(er) tileable textures practice.

I was just thinking of rock sculpts sharpness ( Great pointer Teaandcigarettes !)
Left in these pics is the base version , right is a refined, sharper sculpt.

Done in Blender sculpt , using a standard scrape peaks brush, in 2nd vers. i just took care to set it to cut not too deep into the blobby base shape , and instead widened the range of each flatten/scrape to reach the border of the cut plane and get sharper edges .. but indeed i'm looking into more specific brushes (curve clip and similar)

Darn those sculpts by Ilya Ndyal's are good





Also, i want to share another experiment , on wood glossiness , here the base vers is on the right , left is refined version .. something i wanted to test for some time:

I made the center , flatter part of the planks veery glossy , almost mirrorlike , then lowered to almost black the specular of that same parts... And it works, you don't get much of a metallic look but the intended "glazed"/varnished look..



Below is the maps for 'glazed' version, under that the original version .. check how bright the glossiness channel is for glazed .. but in combination with very dark spec intensity the result isn't super shiny.
As i hoped it'd work... makes perfect sense in PBR terms , but isn't very intuitive as you're making the maps.