Help me refine my workflow when using imported photos in substance designer

Irkie500

I need some advice on how to best breakdown converting pictures that I take myself or source online to tweak-able materials. What I am doing now is this: 
-Import source photo to Bitmap2Material to generate BaseColor, Normal, and Height Information
-Import those Bitmaps into Substance Designer
-Tweak/Make changes as needed
-Save as .Sbs into a library folder for later use inside designer

My biggest issue I am encountering now is that the size of the exported .Sbsar files are massive, and I attribute that to all the bitmaps being used inside the material. So that brings me to a new question, any tips for optimizing the size of the material when I am done? I suppose I could just use the base color map, and then generate the normal from the height map, and use a simplified roughness map.

The source pictures that I am using now from Textures.com are great, but to me they still lack that crisp detail that fully procedural materials do even at 4k resolution. Is that just a downside of using pictures instead of going full procedural? If so whats the point of even using them?

Just trying to work things out so if I need to completely abandon what I am doing thats fine, I just don't want to get too far down a path before turning back.

Thanks!

Larry

Set the size of your first node imputs (the first node you pluck your bitmaps into) to "absolute" or the size of your project in general (just double click on the empty side of your graph, top right is your size options) and change the resolution to what you want, e.g 1 or 2k.

What is the size of your project? A good thing substance designer has (but in your case a bad thing) is that the resolution there is non-destructive. Meaning you can pump it up and down without destroying your pictures. Also wherever you have Grayscale information, you might want to put a grayscale conversion node(or just test it) and wherever you have only 1 color as an image, reduce its size to 1 pixel. Do not use bitmaps for 1 color only, substance designer has color picker so you can just add a 1 pixel uniform color node

Irkie500

Larry said:

Set the size of your first node imputs (the first node you pluck your bitmaps into) to "absolute" or the size of your project in general (just double click on the empty side of your graph, top right is your size options) and change the resolution to what you want, e.g 1 or 2k.

What is the size of your project? A good thing substance designer has (but in your case a bad thing) is that the resolution there is non-destructive. Meaning you can pump it up and down without destroying your pictures. Also wherever you have Grayscale information, you might want to put a grayscale conversion node(or just test it) and wherever you have only 1 color as an image, reduce its size to 1 pixel. Do not use bitmaps for 1 color only, substance designer has color picker so you can just add a 1 pixel uniform color node

Hey Larry thanks for stopping by to try and lend a hand I appreciate it. I am playing around taking textures off textures.com and trying to turn them into usable materials. Most of the pictures I am using are 2k resolution but as much as I try and tweak them they just don't look convincing to me. Since I start inside Bitmap2Material I use the BaseColor bitmap for my albedo so I am not using that bitmap for just 1 color.

As a small test I exported a .Sbsar file with 3 bitmap nodes at 2k(Base Color, Normal, Height), used a HBAO node to pull in the height data, and used a uniform color node set to 16x16 grayscale for roughness and metallic. When I exported it the file size was 11mb! even setting the compression to JPEG on all the bitmap nodes did not help very much.

I am concerned that trying to use that in an actual game would cause all sorts of issues with filesizes that large. Which begs the question if the file sizes are going to be that large what is the point of using those photos and trying to convert them in the first place? I am still learning SD but this allows me to play around a bit easier, I was just curious if anyone else uses a similar workflow and how they get around these issues.

Larry

From the way that you were talking about the sizes i thought you meant something like 200mb... 11mb with 3 bitmap pictures? That is pretty low...! You can edit the bitmaps in MS Paint and save them as JPEG, this way you can save alot. Textures.com is a great place to get samples from, but they are not all tileable. Basically you can take any picture from google and tweak it in photoshop to make it tileable. Making a normal map out of a picture is a good technology, but gives not so good results. If you want to make powerfull textures in substance designer, you should make them from scratch.There are plenty of tutorials to help you get started, both in youtube, and in alegorithmic. I suggest you learn how to produce your own materials and then proceed to making materials from textures. It might seem easier but the results will always feel somehow "off" like in your example. Good luck in experimenting

Irkie500

Larry said:

From the way that you were talking about the sizes i thought you meant something like 200mb... 11mb with 3 bitmap pictures? That is pretty low...! You can edit the bitmaps in MS Paint and save them as JPEG, this way you can save alot. Textures.com is a great place to get samples from, but they are not all tileable. Basically you can take any picture from google and tweak it in photoshop to make it tileable. Making a normal map out of a picture is a good technology, but gives not so good results. If you want to make powerfull textures in substance designer, you should make them from scratch.There are plenty of tutorials to help you get started, both in youtube, and in alegorithmic. I suggest you learn how to produce your own materials and then proceed to making materials from textures. It might seem easier but the results will always feel somehow "off" like in your example. Good luck in experimenting

Thanks Larry! Through some tweaking here and there I was able to get the filesize down to acceptable levels, but like you said the textures always somehow look "off". Been working through some tutorials on substances hoping to get more comfortable with the program.

I think the biggest hurdle is thinking a few moves ahead in how to want your result to look and how to work backwards with the nodes to produce that result. In addition learning how to really use all the nodes together is what makes the program so powerful, but also very difficult to use. Ill keep plugging along!