So I've been watching a lot of the NBC show Las Vegas, and the first season had a lot of people throwing CG dice. And I kept thinking to myself, "man, I could make something more realistic than that". So I put my money where my synisism was and modeled some dice. That was probably easiest part of what turned into a fairly time consuming project.

After the model came the textures, which turned into an oppurtunity for me to familiarize myself better with V-Ray which I've been wanting to do for a while. For some reason it's fairly difficult to find some good tuts on V-Ray, unless you dig around Google and a bunch of forums for a while.

I rendered a bunch of tests of different refraction settings trying to find one that I thought fit the image in my head of what dice looked like. At first I settled on this.

Some things still weren't correct though. Wasn't getting the dots to show through on the other side, and I wasn't getting any color bounce or refraction. I toyed around with some more settings and finally solved those issues. Next I wanted to give my die a proper home.

I decided that I'd make a craps table. I went back to Google to find some references and hopefully a decent texture I could nab. Everything I found was low-res and wouldn't work for what I had planned. So I took bits and pieces from a bunch of images and made my own design. I drew the skeleton vector art in Flash (it's got such an easy drawing tool), then ported that data over to Illustrator where I added the text and sized everything appropriately, then I brought that art into Photoshop where I made the difuse and mask maps. Then in Max I created a plane and applied my texture. I used the mask texture so that I could lay down whatever texture I wanted for the felt. Which unfortunately (for now) was this low-res pic I found online. Here's the result.

Things are definetely shaping up more to my liking at this point. I noticed now that I was using the V-Ray physical camera, and with the lens distortion I'd set up, that the edges of my plane were showing. Now I knew at this point that I'd wanted to change the angle for something that looked more like the camera was lying on the table; so that meant I was going to have to build the meat of the craps table. I knew it wasn't going to be the focus of the piece, so I focused mainly on the shape and the details that'd be viewable in the angle I had planned. That detail was the spiked padding lining the bottom. I didn't want to model every spike, so I used a tiled gradient for the displacement, and whoo-lah, done with that. Also modeled some simple chips to add a little more depth and color to the background.

Now that I had all the pieces, I altered the camera angle for something a little more dynamic and started playing with my depth of field. After messing around with the settings and some test renders, I found I couldn't get the result I was looking for. So I ended up going the old fashioned route, and rendered out a regular pass, and a Z-Depth pass. Brought those into Photoshop and After Effects to see where I could get the best result. I settled with the Affter Effects result.

Oh, almost forgot, I also added a Croupier stick for a shadow that crosses the table.

At this point I know I'm getting closer, but I'm still bothered that I can't figure out the DoF thing for the camera. I've taken plenty of photos (in the real world) and I knew what settings I've used and liked, but for some reason things weren't working. So I did some more reading online and found that it might be a scale issue. To test this theory out I setup a simple scene with some proxy objects that would roughly be the same scale as real-world dice. And whatdya know, it worked.

Now I had to scale everything down in my original scene to be the proper size... great. The easiest solution was making some proxy shapes for the camera and lights, and then linking those shapes along with everything else in the scene to one of the dice. I then merged one of the correctly scaled shapes from my test scene in, and then scaled everything down to the proper size. Then I centered the camera and lights to their proxy shapes, and resized (not scaled) them accordingly. After some adjustments to my camera settings and some test renders, I was in business as far as DoF goes.

After some further scrutinization I noticed there's a slight lip to each divot on the dots of dice, which my model was missing, so I added those edges. Also, I noticed the surface of dice is a bit more broken up when it comes to reflections, I'm guessing due to finger prints and scratches. So added some subtle noise and dent textures to the reflection and bump maps. And finally after some more searching I was able to find a higher-res felt texture. I applied that, and ditched it's old low-res bump texture and opted for a noise map instead. Lastly I added a low level light so that the faces of the dice that were towards the camera would have a little extra gloss on them. Now it was finally time to render a high-res image, so I hit the magic button and went to my living room for some chill time.

The render ended up taking 8 hours 6 minutes. The next day I realized a couple things. One, that I could've saved the irradiance map which would've made any future renders a lot quicker, and two, that I had forgotten to turn 'render iterations' back on my dices' turbo smooth modifier. Since I didn't save the iRad map and that last render was 8 hours I decided to let the dice-dot edging slide for now.

When I decided to make this an actual scene and not just a pair of dice, I knew that I wanted to have some kind of title piece that'd fill in the empty space next to the die. So during this process that thought had been bouncing around in my head and I decided that I was going to go with the infamous Vegas sign, with a slight twist.

Finding some good refs of that sign turned out to be more difficult that I anticipated. There were plenty of photos on the net, but not many showed much detail. For example, I could tell that there was something written or drawn under each letter of the 'welcome'. After some research I found out that each one of the circles represents a silver dollar. At this point I was guessing that the markings I saw was the face of the dollar, but wasn't sure. I finally came across an image on flikr that showed enough detail of random areas of the 'welcome' for me to determine that it was in fact the face of a 1922 silver dollar. After all the research was gathered created the textures much in the same fasion as I did for the craps table. But this time I was able to use the splines I drew for the texture in Max as the base for my model. Like for the neon tubes, and body of the sign, for example. When the model was built and textured I was left with the placement of the light bulbs that surrounded body of the sign. Now I didn't want to place each individual bulb around that sign, luckily I found a hidden tool called 'spacing' .. or something like that, can't remember right now. It's in the 'array object' dropdown if you want to check it out. But it basically allows you to select a spline path and space out as many objects as you want along said path. Then I grouped the instanced lights together and threw a cylindrical uvw map on and scaled it to fit around all the lights. I then added a checker texture to it's multiplier slot so that every other light would be on or off, and bam!

I was now in the final stages of this project. And I knew that I didn't want another 8 hour render, so put together a seperate scene for the sign. It had all the same models, cameras, and lights; except I used V-Rays version of a matte/shadow map on all the objects I didn't want to render. What's cool about V-Rays matte/shadow is that you can assign a base material that'll still reflect light and color the same as the original texture, but it'll render with a black alpha channel, huzzah. After some test renders I decided I wanted to add a base to the sign to make it appear more as a souvinir and also add a couple tags to the sign face. The first is my sig with an umlaut above the 'A' to give it that 'Las' Vegas sound. And the second, I painted over 'd-a' in Nevada, and replaced it with "dull". Then I rendered her out along with an object channel for the dice. I needed the object channel because for some reason the reflections would show up on difuse pass but they wouldn't show in the alpha. That render only took an hour, huge improvement.

Now that I was almost done I decided that I would re-render that section of the dice that was bothering me. Luckily since this was a still and not an animation I was able to do a region select and just render out that bothersome section. I think that section took about an hour or two, damned refractions + DoF + my ignorance of iRad maps.

With all my ducks in a row it was finally time for Photoshop. After stacking some layers and a mask/blending change for the sign reflections, I was done.