As promised in the latest post of this series , here is the final model of our game character called Armin, a mythical Persian warrior who clobbers his enemies with his two gigantic maces.

This character was modeled and shaded by Reza Ramezani using 3DMax 2009 and ZBrush 3.1based on concepts by Faraz Shanyar.

Armin warrior character concept

Armin final closeup

Armin final model

The model was textured in Photoshop and rendered using Vray and the whole process took approximately eight working days (we weren’t working on it full time). Most of the process was pretty straightforward except a few problems we had with Vray’s sub-surface scattering material. We were trying to obtain a rather realistic but at the same time stylized look for the skin of the character but we kept bumping into limitations on this shader.

Armin warrior ZBrush model

We had never done sub-surface shading in Vray before and we didn’t have any detailed documentation on the workings of the sss shader so we tried all the parameters and finally decided to use a blend material with the sss shader in one slot and a vray material with the paint map in the other. It worked pretty well and we reached 80% of the result we were trying to achieve and decided to add the rest in post. To do this we rendered another layer of skin using Mental Ray’s Fast Skin shader which gave a very good result and we added this to our character in our Fusion comp.

Here you can see some of the texture maps we used for this model:

Some of Armin’s textures

And the lighting of the scene which consisted of two spot lights, one acting as a key light and the other acting as a rim light. The rest was environment lighting and environment maps.

Armin scene lighting setup

Here you have a screenshot of our render settings but we didn’t really fine-tune these since the render was pretty fast and we were getting a decent result:

Armin’s Vray render settings

These are some of the render passes we used to composite this image. We hope we’ll have the time to rig and animate this character pretty soon and we will give you an update on that as soon as it’s available.

Armin render passes

We are currently working on a scene which will contains a large field of flowers and vegetation. Currently our effects expert, Adib Souly is working and testing different scenarios in Softimage XSI. You can see a simple concept we did (this was done in a rush by Faraz Shanyar) that was done to help us envision the density and distribution of the vegetation we would need.

But the problem was much thornier than we expected …

We knew from the beginning that there was no way in hell we could create and render such a huge scene which contains camera animation, animated shrubbery and vegetation and millions of polygons without some optimization. We also had a model of a flower we had tested and needed to implement in our scene in numbers as you see in the concept. One other important factor, which was very important to our timing, was that the solution had to be scalable and portable since we had a total of at least 25 scenes which would need to implement the solution and the mere idea of doing it all by hand was painful.

Our first solution was to use Softimage XSI hair in different layers that would constitute our levels of detail. These LOD’s would be created by using a mathematical expression that would link the density of our hair to their distance from the camera. Unfortunately this did not give a very good result in our tests as the number of hairs would not decrease or increase smoothly. Instead it created a sort of procedural animation that we feared would be noticeable in our renders.

The next solution was to use hair again but instead of animating the hair density we decided to create a multitude of hair patches which would have decreasing numbers of hair strands and their position with respect to the camera would be controlled by weight maps which needed to be done manually for each scene. This could obviously create some problems with some of the scenes where the camera animation was more intense but we figured that we could eventually used animated weight maps if the need ever arose.

We decided to use lightweight objects as instances on the hair patches for the layers that would be near the camera and simple hair strands on layers that are further out.

We still had problems with rendering our scenes, we would run out of memory or crash XSI or Mental Ray all the time so we knew we had to look for a simpler solution.

We even did a test on E-On Software Vue 6 but the frame took too long to render and we didn’t have enough experience on Vue to be able to optimize it in time.

The next step in our optimization was to opt for textured sprites of rendered vegetation that would always face the camera still using the hair system. The advantage of such a method would be that we would use much less polygons and even lesser hair strands. In fact, we are still testing this stage and haven’t reached a final solution but the inconveniences of such a solution would be the possible low quality of sprites near the camera or their 2D nature in extreme camera movements. There is a white paper done by Pixar on scene simplification and optimization called Stochastic Simplification of Aggregate Detail which goes into great detail regarding this very problem, but the truth is that this solution was too technical for us to implement and we’re not even sure that it could be easily implemented in XSI due to it’s procedural nature.

For the moment we haven’t reached a verdict yet, but I’ll post our results as soon as we have any.