Inspirations
I enjoy 3D modelling, and my style of choice is a quality cartoon style something like these:


As I develop, I would like to be able to produce more complex models and I have therefore chosen Wizard Cat by Maria Panlilova and Decentraland Medieval Plaza Scene by Fabian Orrego to research and base my development targets on.
Wizard Cat
Wizard Cat was inspired by Japanese and Korean fantasy art and was created by Maria Panfilova, a Russian 3D character artist. In creating it she used ZBrush, Maya, Substance Painter, 3DCoat and Photoshop. She is strongly influenced by the work of Ryota Murayama and spends a lot of time studying his and others’ work as references. She uses these studies extensively to enhance her own use of form and colour. When creating a model, she uses a combination of her memory, photo references, and other artists’ work for each body part.


When making Wizard Cat, Maria chose to build a high-poly model and bake it down to a low poly as she found that this method gave more consistent results that were closer to what she had designed in ZBrush.
ZBrush
Maria’s original plan was to make a party of three characters, an archer, a mage, and a heavily armoured character, all in an “epic pose facing danger”. The archer was to be a cunning fox-like creature, the mage a cat, and the heavily armoured model something like a bear. This led to her first sketch:

Concentrating on the cat mage, and starting with ZBrush, she made 2D concepts, switching to a flat shader to allow her to ensure she was happy with the silhouette. This is the most critical part of the ZBrush process as it is used to make a low-poly.

Using the “mask by polygroups” functionality in ZBrush, she was able to paint different mesh parts separately from each other without needing to make a clean detailed sculpt.


Using references from Sketchfab – turning on wireframe mode and viewing from all sides – helped her to perfect her wireframes.
By making the staff and book as separate objects, they can be replaced by different objects during game play.
A trick she used when baking her models was to separate overlapping objects into groups, making a single frame animation which separated them from each other before the baking process. This made sure that the meshes wouldn’t bake weirdly due to overlapping.

Substance Painter
Substance Painter allowed her to set up the lighting angle and colour temperature (warm or cold), combine it with diffuse and reflected light, and most importantly to be able to switch lights on and off and see the effect immediately. Even though this is somewhat rough and ready, it provides a good base for further painting. Setting the position map to One Axis makes gradients from the top to bottom of the model, which helps to create an effect which draws attention to the top part of the character – an effect similar to a vignette in a photograph. Gradients can be in luminance (light to dark) or colour temperature (warm to cooler). A combination of these is used in Volodya Liubchuk’s model to lead the viewer’s eye to the head.

By setting different ID masks for each material, Maria was able to bake different lighting settings for each material. This allowed gold to look shiny and metallic. The result before any painting was like this.


Painting
Using 3D-Coat Maria then painted over her net. She used this program because it provided a handy bridge to Photoshop so she could carry out some actions like colour correction and detail in Photoshop while carrying others out in 3D-Coat. 3D-Coat also allowed her to merge layers to simplify her workflow.
Scrolling through references of hand painted textures from other artists and allowing them to influence her style, she then added highlights and shadows matching her paint colours to the overall lighting colour balance – warm shadows and cool highlights.
Pose
With the texture almost completed, she switched back to ZBrush and imported her lowpoly model to finalise the pose.
Ground
Using a similar process, working with ZBrush, Substance Painter, 3D-Coat and Photoshop, she made the ground before making final tweaks to the texture and combining the parts.
Decentraland Medieval Plaza Scene

Decentraland Medieval Plaza Scene was created by Columbian 3D artist, Fabian Orrego using Blender and Photoshop. He used Blender throughout for every part of the process except the original concept art (for which he used Photoshop) and the final presentation (where he used SketchFab).
Concept
While not completely necessary in all projects, having a concept drawn out made the modelling process much easier. It allowed Fabian to quickly produce ideas in 2D where they could be modified more easily than was possible once they had been created in 3D.
Using reference images from other games or concepts (e.g. Pinterest, Artstation, and Sketchfab) he created line sketches of the plaza before making the final coloured sketch.



He then moved on to 3D modelling.
Blocking

Fabian blocked out a model as a starting point to help him scale the elements of the plaza as a whole. This allowed him to make adjustments to the content before adding detail.
Modelling and Texturing

Adding detail to the block model, he started with the largest object, the castle, adding progressively smaller objects like the stable, the farm, etc. Once all the buildings had been made, he added the terrain and vegetation, by sculpting the blocks from the original block model. Subdividing the model into a number of areas to work on individually helped him keep track of his work.

To ensure that the relative sizes of the buildings were correct he used one of his character models as a guide.

Careful labelling and layering ensured that he was able to track each component and he took care to use as few polygons as he could without compromising the style.
Moving on to texturing the model, Fabian created a colour atlas of gradient colours in Photoshop. This approach allowed him to texture large areas with a single texture, accelerating the process of UV mapping.

Rigging and Animation
Any animated models in the scene were then rigged. Careful planning allowed him to minimise the number of bones to rig and facilitate naming each Armature in Blender. In the scene he animated objects that would naturally move – water, trees, flags, making them move in a way consistent with a chosen wind direction. Because all the animations operate in a loop, he first defined an overall global loop. All animations had to be this time of a smaller ‘sub-multiple’ of it, e.g. if the global loop was 300 frames, animations could be 75, 100, 150 frames, etc.

Organisation & Colliders
The final part of the construction was to organise the scene into zones, each in a different layer in Blender, ensuring consistent and correct naming. Each zone was allocated a colour to make it clear.

The original blocking was then used to define the colliders which were used to ensure that the user could not pass through “solid” elements of the scene.
The finished scene was then published on Sketchfab.