Fatih Gurdal
Game Artist

Home | Back

The first thing I think is worth mentioning is that I have made two base meshes for humanoid characters.
One for males and one for females. These base meshes have some detail sculpted already, but wouldn't consider them "final".
I consider these base meshes to be a "sketch" that I use to base my "line-art" off.

The first thing I do when making a new character, is to make sure the proportion of the model fits the concept art.
On the left is one of the base meshes I mentioned earlier and the one on the right is a quick and dirty edit to fit the new proportions.
At this point I am not too concerned about the quality or details of the edited mesh, as the concept of the character has her whole body covered up.
The only part I detailed a bit was the face, since the bottom half of the face is the only part of this character that is exposed. =]

I consider the head of a character to be one of the most important parts of a character, so I started working on the head first.
It can be tricky, with hard surfaces, to retain a nice smooth surface while trying to model in some details.
Because of this I decided to use a method I like to call "reverse retopology". Basically with this method I completely ignore the shape of the model and focus purely on the topology first.
After I am satisfied with the base topology, I subdivide it once and use an FFD modifier to give shape to the flat surface.

With character models, since I already have base, I basically just draw polygons on the surface of the body to create the gear. This is basically the same as retopology.
Once I have analyzed the concept art, i take not of all of the details, model a few key details and copy and paste it where needed.
The concept art that was provided for this model did not have a back view, so I had to wing it. Again, duplicating existing models, to save time as well as keep a consistent look.

When creating these details, I try to avoid 90 degree angles. There are two main reason I do this.
The first one being that when baking the high poly to low poly, it works better for the normal map.
The second reason is that it will catch specular reflections better, making the model read a little bit better.

Here's an example of how it works. Basically with sloped angles, especially when you have low resolution textures, the normal map reads better.
The version on the left has the sloped angles modelled into the high poly, the version on the right has no/very small bevels on the edges.
It goes without explanation the bevels on the right are almost non existent, when the normal map is applied to a plane.

Here is the final high poly model and base for sculpting, for organic surfaces like cloth and leather.

Final sculpted high poly. Some pieces are missing as the will get mirrored.
This is done to save on texture space as well as time.

Final high poly model with the mirrored parts.

Next up is retopology, making a game ready model from the high poly base.
I like to start out with a modelling application like Maya to layout the overall topology first and model repeating parts, like the fingers, that I can copy and paste.
After I have the base topology done, I go in and start adding detail so the low poly mesh reflects the high poly more closely.

The first pass of the low poly model, it ended up a bit lower than the given specs. (4000 triangles instead of 5000)
Before I started adding more detail I made a quick auto unwrap and baked the model, the check how it would bake.
Overall a pretty good bake, not much artefacts.
With the model and the test bakes I evaluated the model to check what areas could use some more polygons.
I ended up decided to smooth out the silhouette of the model so it looked less blocky.

The final low poly model, sitting at 5000 triangles.

I deleted parts that would get mirrored on the UV unwrap.

With he model now unwrapped, I did a final texel density test and made another test bake.
Making use of a shader that has a Fresnel effect in both the diffuse as well as in the self illumination channel, I check for any artefacts that need to be fixed.
I use the diffuse Fresnel and illumination Fresnel as it allows me to "read" more information from the model, as opposed to a regular lambert/phong shader.
The test bakes came out with a satisfying quality, so they are used as the final bakes.

When it comes to texturing, the first thing I do is make masks according to the materials, as seen on the left.
Using masks like these allows me to quickly select and/or modify everything that should have the same material properties.
With the base colors laid out I mix in my bakes. Ambient occlusion, cavity, lighting, etc.
The map on the right is the result of all the mixed maps, the base colors and a detail pass.
The detail pass is mainly for adding wear and tear, adding low key details to make the model look a bit more interesting.

Next up, I make a specular map based on the diffuse map. Since I had separated all of the materials using masks, I could simply select a mask and modify a set of materials with ease.
Again using masks for the individual materials, I created a gloss map.

On the left is the final normal map, I haven't edited it much. Went for a "clean surface" look and kept all the detail on the diffuse and specular map.
The map on the right is a low res(256*256) glow map, for all of the glowy bits on the characters.

And the final result:

Top | Home | Back