Finally got around to putting something together for how it's done

Typically the process starts with modeling. Every object in the scene must be modeled, no matter how small. This is time consuming and a challenge to get the shape just right. It usually takes me about half the time for the total project. It is very much like modeling an object in CAD for design or manufacturing but without internal parts (unless you can see them). Basically, you create a wireframe of the object using a number of tools and views. Also, in this case the sky was painted in Photoshop. I don't use photos, I've just never liked the look of a untouched photo, in an art piece, it just looks out of place to me.




Next, the objects can be placed in a "scene" which is a virtual studio with lights, camera and tools for controlling the look. This is where you set up the composition of the piece and unlike painting, it can be changed (a good thing for me!). Setting the lights is one of the most important steps. The scene is also where I do simulations for smoke, fire and grass.


Up until now there are rough shaded previews but at this point you can "render" a much more accurate view. The computer calculates the look based on your lighting. The rectangle in the middle is "rendered" at a test resolution.

In parallel with this process, I work on the "textures" or paint job of the object. Here is 1 layer of the tank camo.

Setting lights only goes so far. In a scene like this, HDR lighting is the best. It is a high dynamic range image of the environment wrapped around the entire scene. This is a panoramic image shot at multiple fstops to capture the entire dark to light range and this is used to actually project light into the scene (see the sphere surrounding the scene). This gives nice bounce light to make objects fit in the scene naturally and take on some of the color of the environment.

The yellow "blobs" in the scene are smoke particles and are simulated with a simplified fluid simulation.

All objects in the scene must have "materials" attached that makes their surfaces look realistic. This controls how light illuminates or reflects off of them.

Textures must then be projected onto those objects. It is a little like using a slide projector to shine a picture onto a surface. This is easy for simple objects but much harder for complex ones. Sometimes objects have to be "unwrapped" to be painted properly. Small curved details take time.


Grass is grown with a "hair" simulator. The length and density are controlled by painting a "weight map" on the object that has the hair.


After all final adjustments are made the scene can be "rendered" at full resolution. A complex scene like this must be rendered in layers and parts because it's too much for my computer (or almost any computer). This is where the computer calculates the look based on everything you've done. It can take several days (this one took 4). So as you can see, I could use a computer several thousand times faster

This part can be time consuming and frustrating at times. It's a bit like playing the piano and hearing your song the next day. It's not always what you expect at the end! Things always have to be redone a few times it seems.
At this point a few extras are made. A depth map allows for depth of field blurring to be added and adjusted after rendering. Darker areas are further away.

An ambient occlusion pass calculates how ambient light is cast on objects. These additional elements are blended in Photoshop.

So, that's it in a nutshell. Hope that explains the process just a bit. Feel free to post questions about any part. Pretty geeky stuff!
