To learn from it I started my 1st simple scene (attached to this post).
It is a simple plane unwrapped, just bend a little bit. I added a simple diffuse material to ut with an image texture. (just white only)
The first render I did is using the cylces renderer and I come out with result_1.png:
Then I changed the image texture back to the 1st one. I saved the render result as OpenEXR multilayer and
used the node setup provided in the tutorial.
I tried to replace the white image texture with the colored one without re-rendering the scene, just by using the OpenEXR file and the provided node setup.
What I got is not bad but not exactly what I wanted to get. Please see the right image.
There you can see the node setup has replaced the texture, but the letters (and the color stripes as well) are smoothed out for some reason.
1st question:
Why the letters are smoothed out? I want them to appear as sharp as in the left image.
2nd question:
In the right image on the right edges of the plane (in the blue area) one can see a violett border, which does not appear using the ‘normal’ rendering. Where does this border come from and how to get rid of it?
My best guess is that the image representation of the UVs doesn’t provide enough information for it to distort the image to such steep angle.
It uses pixels to do that and more pixels = more information. You rendered it in 960x540 (50% of 1080p) resolution, so rendering it with 100% produces less blurriness. Rendering it @200% of 1080p and scaling it back to 100% pretty much clears it, but that’s not really the solution.
I did a higher resolution texture image (with crisp text outlines), and also fixed the diffuse pass by adding the image texture color to RGBx0.01 in cycles material and rendered again. The purpose of that material fix was to get rid of 0 values which looks total black but isin’t, giving a clean diffuse pass. That helped to get rid of the noise in the resulting image but the blurriness is all because of map UV distortion.
Bartek also had to do a workaround for his scene in the tutorial. If I remember correctly, he rendered the box colors in blender render.
Map UV node has alpha value. Put that to 100% to get the corrected alpha for the mapped texture.
Also, the use of switch nodes is not necessary anymore. Reroute nodes can have a label and those are much more clearer in the node layout, although you might need to adjust theme colors. (You could replace them with reroute nodes easily by using node efficiency tools addon, which is made by Bartek).
thanks for your help, I tested it with 1920x1080 and indeed the image gets clearer. And also setting alpha on UV Map to 1.0 solves the violett border problem. Thanks.
In your anwere you wrote:
I did a higher resolution texture image (with crisp text outlines), and also fixed the diffuse pass by adding the image texture color to RGBx0.01 in cycles material and rendered again. The purpose of that material fix was to get rid of 0 values which looks total black but isin’t, giving a clean diffuse pass. That helped to get rid of the noise in the resulting image but the blurriness is all because of map UV distortion.
Could you please explain this in more detail what you exactly did? What do you mean if you say ‘to get rid of the 0 values’?
What 0 values you mean and what problem do they cause?
Could you provide a screenshot of the new node setup you did?
You said:
although you might need to adjust theme colors
What do you mean exactly? You mean I should change my blender theme colors or what?
The texture image has text written with full black color RGB(0,0,0) and this causes the diffuse direct pass to have that too (bottom left corner in the screenshot). I changed the material to add texture colors to RGB(0.001,0.001,0.001) and that fixes the diffuse pass (bottom right).
Now when diffuse direct and diffuse indirect are added and then multiplied with the color in the compositor, multiplication happens on a clean diffuse pass.
Multiplying any color with 0, you get 0. Without the fix you would get that “TEST TEST” text, no matter what image you would use as a color.
With the fix, the original combined pass gets non-zero values for the text (bottom right in the next screenshot) but for human eyes, that is still as black as RGB(0,0,0). (The difference is as invisible for other colors that got the same 0.001 added to them).
Yes, node labels might not show up very well without changing their colors from the theme settings. With reroute nodes, dark or black label color would be a problem since they don’t have a background like normal nodes have.
Hi JA12, thanks for making it clear to me! By the way is there any liuterature available to learn more about render passes and such stuff? What do you use to learn more about it? I think this techique using render passes has big potential, but I think also it is not well known to the community. It offers great flexibility without re-rendering, thats great. So how can I get more in-depth knowlege? Thanks!
You’re welcome. I wanted to say “no problem” but as you probably noticed, there were problems with the forum.
I don’t actively seek more information about it since 3D/Blender is just a hobby and I can’t give much tips on learning resources.
Bartek is the only one I know who produces tutorials that has enough knowledge about the theory, history that leads to that theoretical information, and practical use of anything related to compositing and can break down concepts and issues to understandable form. In addition to cookie tutorials, he has some tutorials on his site and blender e-shop has http://www.blender3d.org/e-shop/product_info_n.php?products_id=161 this. It doesn’t use Cycles though and Blender render passes are more painful to use for compositing. It does contain other information too like color profiles, color correction, color management and such.
I’ve used Gimp and Photoshop for a long time and that has helped to get into Blender compositor very easily. Although Blender compositor is very powerful in a different way. I’ve also done machine vision development in my work and what I learned there about image processing probably has shaped my thinking process.
When you think about “passes” (in any program, by any name …) in the 3D Graphics context, there really are two very-pragmatic labor-saving principles at work here. I think that, if you keep these two ideas front-and-center, most of the rest of it will follow quite naturally. Here goes:
(1) "IT TAKES TOO DAMM LONG. :eek: YOU CAN’T AFFORD TO START-OVER. YOU’VE GOT TO TACKLE THIS PROBLEM BY MEANS OF “SUCCESSIVE REFINEMENT.”
It simply isn’t an option for you to “mash the big red Render button, wait 37 hours or so, then see if you have anything-at-all to show for it.” You need to be sure that whatever you’ve spent all that time producing, is usable, and that it is usable as a step forward towards, (heh…) “whatever the client will finally accept and pay for.” Therefore, you must adopt the same approach that is customarily used in creating any pop-song: (a) capture the various components individually, then (b) “mix-down” to produce the song(s) that will finally make it to the radio. Shania Twain routinely produced both “country,” “pop,” and “dance” versions of all her stuff: three-times the mechanical royalties, but only one visit to the studio. (Studio musicians are expensive, too.)
(2) "EVEN WHEN YOU’RE ASKING THE COMPUTER TO DO “A LOT OF WORK ALL-AT-ONCE,” OFTEN THE BEST WAY TO DESCRIBE WHAT YOU WANT IS TO “BREAK IT DOWN.”
The “Big Red Render Button” is, IMHO, mostly there “for marketing reasons.” Instant Gratification™ and all that. :yes: But, most of the time, the best way to describe a problem to a computer (in such a way that you an debug the thing …) is to break a bigger problem down into smaller problems, then describe how you want the various smaller-problem solutions to be combined. (This is exactly the same idea, “in the small,” that I was talking about in #1 “in the large.”) This is why, if you’ve been resisting looking at “node-based things” in Blender, the time to stop “resisting” has come and gone and gone and gone. :yes:
Therefore, you must adopt the same approach that is customarily used in creating any pop-song: (a) capture the various components individually, then (b) “mix-down” to produce the song(s) that will finally make it to the radio. Shania Twain routinely produced both “country,” “pop,” and “dance” versions of all her stuff: three-times the mechanical royalties, but only one visit to the studio. (Studio musicians are expensive, too.)