Welcome to the second part of the “behind the scenes” look the lengthy process involved in making the upcoming render I titled “The Customer Is Always Right.”
In the first part, I showed you the various steps necessary to set up a fifteen character 3D scene. Today, I will show you the steps involved in making the actual render, and composing the renders together in one image via postwork.
Let me be blunt. The steps you are about to see are steps that I usually do not take. I prefer what is called a “clean render.” That means, everything, including any special effects such as magic or smoke, are all part of the original render and handled through props and textures. Many other 3D artists, however, will run their final image through Photoshop or GIMP and add smoke, fog, or change the colors. There is nothing wrong with that, I just prefer not to do it. It’s a leftover from my years working with a film camera (if I couldn’t see it through the lens, it didn’t exist).
The image below is a clean render. Yep, even the streaks of sunlight are a rendered prop, courtesy of the Reality 2.5 plug-in. This is called “Afternoon Noir.”
“The Customer Is Always Right” forced me to abandon the clean render, and decide on methods to optimize postwork (the process of enhancing the image in an editing program).
At this point, I need to explain a little bit about the figures used in this scene. I’m using the Victoria 6 and Michael 6 figures. They are the latest in 3D figure technology, and appeared in 2013. These figures (and the ones spun off from them) are collectively called the G2s, for Genesis 2 figures. The most G2s I’ve ever been able to successfully render in one scene has been seven. You see the results at the top of this blog. Those seven only fit without freezing my computer into a glacial state because I used every trick I know, such as using a prerendered image for the background, and making all body parts hidden by clothing invisible, so the computer doesn’t try to render them. “The Customer Is Always Right” requires fifteen G2 figures.
Granted, I once made a clean render with thirteen figures in it. It’s called “The Characters Escape.” Those figures, however, were all Victoria 4 and Michael 4 characters. That technology reigned supreme from 2006 until 2011.
Let me show you the difference between Victoria 4 and Victoria 6.
These are the wireframe meshes for the basic “out-of-the-box” figures. Victoria 4 is on the left, and Victoria 6 is on the right. I bypassed Victoria 5 and Michael 5 (the now almost abandoned Genesis 1 series). Notice how much more realistic Vicki 6 looks compared to 4. She has more polygons in her mesh than her older counterpart. More polygons means a more rounded figure, and humans are a collection of cylinders and spheres, not rectangles and squares. The increase in polygons is more apparent as we travel down the body.
Again, Victoria 4 on the left, with Victoria 6 on the right. You can imagine the difference between their torsos (as the ladies are nude, I won’t embarrass them). This increase in realism comes with a price. The more polygons present in a mesh, the more RAM and CPU resources it uses. Ergo, I can fit thirteen 4s in one scene, but only seven 6s.
This meant I had to split this scene at least in half. I studied the screenshot of the assembled cast in the set for an hour, determining the best way to divide the scene. For the sake of rendering time, and ease in combining the pieces together, I chose to split the scene into four parts, based on overlapping characters and shadow fall. The divisions were right, right center, left center, and left. Each separate scene featured five characters, with at least one appearing in two scenes to provide overlap and a fixed point of reference. My preference for giving each character his or her own file, and merging them into the set, made this process easy.
Of course, that’s theory. In practice, I had no idea if I could even pull this off. The scene may just be too complicated. In order to find out, and to check the various textures under actual render conditions, I made some test renders of all four groupings.
In case you’re curious, here is a screenshot of the setup screen in DAZ Studio for Image One, Group Right:
Below is the Reality 2.5 console. This is where I make metal look like metal and wood look like wood. It is also where I launch the actual render, and send all this binary information to Luxrender to turn it into an image:
I’ve mentioned Luxrender before. Below is a screenshot of the image actually rendering:
It looks like a grainy mess! Over time, it will clear until it is as sharp as a photograph. I don’t need to let it run that long, just enough to check textures and test the feasibility of composing four separate images into one. Notice the controls on the left of the screenshot. Luxrender simulates the effects of actual film and cameras. In my case, I use Advantix “film” for its more subtle coloring. The Fuji simulator gives more bold colors while Kodak offers a more golden look. The ISO, aperture, and f/stop are for those of us who are accustomed to camera response. I set the camera controls at ISO 100, f/stop 11, and shutter to 1/60 because that’s what I would do to my Nikon for a shot like this.
These renders were all made with the same specifications: 75% size (I usually make a render at 200% size), the exact same lighting settings, and the exact same rendering time of 250 s/P. That’s Luxrender talk for 250 samples, or the number of passes the render engine makes over the image to clarify and sharpen it. Bare minimum is considered 500 s/P, and many good renders reach 1000, or higher. Oh, heck, I don’t know what it means! Remember, I’m the man who insists that two plus two does not have to equal four. I just know that high s/P means a really sharp render. 250 sufficed for tests. 250 s/P makes a better point of reference than the time required for a render to run. Time is affected by textures, so a scene with a lot of metal will take more time to reach 250 s/P than one with no metal.
Image One, Group Right:
The first group looks good under lights. Issues spotted: the Customer is wearing leather boots, but they aren’t shiny enough. The lutenist has the opposite problem; his clothes are too shiny. The woman seated is reaching for her sword, but we can’t tell that’s what it is. Her hair also looks a bit too orange. The clay tankards overhead are way too bright. The tankard on the floor isn’t hollow. I’d like to rough up that floor a bit, too.
Image Two, Group Right Center:
Overall, this looks really good to me. The dagger could be a bit shinier. It also casts a shadow that could present a conflict with Image One. I may need to include just the dagger in Image One so its shadow falls behind the Customer’s leg.
Image Three, Group Left Center:
Uh, oh. Some major issues. First, the couple in the center casts a shadow on the beam to the right of the door. In the previous image, we notice that a woman stands by that beam. That means, I need to break this couple into a fifth group that includes that woman, so I have their shadow upon her. The woman in the center is about to draw her sword, if the redhead at the bar draws hers. The action isn’t obvious. She needs some reposing. The man holding the tankard is wearing cloth, not leather. Turn the glossiness on those clothes down! His skin looks funny, too. He’s blue, so I will check his skin texture settings. Something you don’t see in this render is that his hair caused a critical error warning in Luxrender. I’m not sure what the problem is, but that sort of error can crash the program. So, he needs new hair. Finally, the fiddler is just one big disaster. I tried to pose her as if she was just knocked down by the fallen barbarian, and in trying to rise, gooses the lady in the red dress (see the next image). Her pose doesn’t work. Also, that fiddle just plain sucks! Time to find a new prop.
Image Four, Group Left:
I already covered the fiddler and the man in the back. His wife looks too dark, and it isn’t shadow. Her skin texture settings are too low. The man on the left has it the worst. He’s supposed to be half of a husband-wife mage couple. She looks like one. He doesn’t. He needs a new costume, and I think I’ll switch his clothes with those of the lutenist. His hair looks flat. Also, that tankard is supposed to be metal. I am also tempted to retexture the red dress into something more original.
Now, I leave the world of 3D rendering and move to a world more familiar to many of you: digital art editing. I use GIMP 2.8 because I can’t afford Photoshop. It does the job. I won’t describe the layering process. If you’ve worked with layers, you know what I’m about to do. If not, it’s too involved to explain here.
Armed with the four images above, plus Image Five (the couple casting a shadow on the woman by the wall), I set out in GIMP to see if they could be merged together into one complete image. In a new screen, I arranged the images as layers in the following order from bottom to top: Image Two, Image Five, Image Three, Image Four, and Image One. Then, with a tremendous amount of tedious and precise erasing, I arrived at this:
I feel like I should wander around the apartment like Baron von Frankenstein, shouting “It’s alive!”
So, after making the adjustments listed above, running some more test renders at the same settings, and adding a sixth layer to kill the glare on the rafters from the lights, the result is …
The way the image looks now, it needs a new name. I’m thinking “Just Another Saturday Night,” based on my favorite character, the guy in the center, leaning back and enjoying the ruckus. The next time you see this image, it will be finished and full size.
Is it art? My personal definition of art is “the telling of a story through the written word, the spoken word, visual imagery, music, drama, or physical material.” So, yes it is art.