Preserving the style on avatars and their original PFP. Case Study: Regulars
This is some text inside of a div block.
This is some text inside of a div block.
Nov 8, 2023
We converted select characters into VRM format, capturing the 2D aesthetic from their 3D counterparts, inspired by Fernando Botero's paintings
When turning a PFP 2D image into a 3D interoperable avataryou must take some considerations, not only optimization wise to make sure that the avatar is compatible across realms but also to make sure the style is preserved when creating the new parts that are unseen from the single render available.
One of the latest avatars requests we had to adapt to 3D came for a private owner. 3 avatars from the collection Regulars created a pretty interesting challenge: how can we could maintain the unique art style from the 2D image to the VRM model.
To give some context, Regulars are an experimental project focused on online personas, character development, and on-chain assets. And it's a fantastic opportunity to visualize how this unique art style can be transported to full body identities taking as an inspiration the famous Colombian painter, Fernando Botero.
Botero had a very characteristic style which makes him a timeless artist whose paintings won't leave anyone indifferent. His artworks often represent voluminous characters and explosive color palettes. These elements are often used to represent political criticism or humorous scenes, much like Regulars, whose designs are undoubtedly extraordinary.
With everything in mind, our objective was to create a 3D avatar that reflected the faces of the Regulars while we widened to new and unseen areas like the body, taking some references for body shape and painting style from Botero.
Maya or Blender
Step 0: Pre-development
We had a phase in which we took some references from paintings for the body shape and face.
We can appreciate how the painting reflects love not only for the volume, but the body is clearly shaped with spheres, giving a good cohesion across all body parts.
Just as the style shows how the face has very small proportions and the facial features are far apart from each other.
Step 1: High Modeling
We begun our design by sculpting a body that worked for the VRM design, taking the lead with Trisha. As we progressed with her female shape we aimed to keep the round shapes to make a steady connection with the face in the portrait.
For the face, we kept its similarity to the reference as closely as possible.
Adittionally we included some of the creases that Botero added to his models, such as those in the hips and waist. You can compare some of these shapes in Botero’s paintings of the dancing couple.
The hair presented an additional challenge due to the mesh limitation of triangles (tris) in the VRM format when aiming to create an avatar that is optimized for any platform. We decided to model the hair by dividing it in big sections and adding details to those shapes.
After we finished the main body, we had to dress it. For that purpose, we decided to keep the clothes fitted to the body in order to showcase the shapes we had created. As Botero described his works as exploring the “sensuality of form”, we followed through to ensure the style was preserved throughout the entire avatar.
Step 2: Low resolution & Textures
Once we had the high definition model, we created an optimized version with no more than 10.000 triangles for the entire avatar. We aimed to keep the model as low-poly as possible, allowing it to be used in different platforms and easily carried with you.
Regarding the shader, we chose the PBR (Physically Based Rendering) pipeline. This decision allowed us to keep the lighting data enriched and keep the normal map data when the model received direct lighting. With this, we would put detail on the textures instead of putting it on the model mesh. With the use of the Roughness, Metallic and Occlusion channels we gave the model all the required details to make it look like an art piece, just as the original PFP image.
Step 3: Rigging and Blendshapes
The blendshapes in this project were kept discrete and exaggerated facial expressions were avoided. We aimed to keep the blendshapes realistic since one of the mottos of Regulars is that they are common folks. To fulfill this objective, we limited the expressions to fit within the character’s shape, keeping the mouth relatively small.
Below, you can see the different blendshapes we created for Trisha.
When we approached the rigging process,we used the standardized humanoid rig of the VRM format. However, we made the decision to maintain the same bones, in terms of position, rotation and proportion. This not only to speed up the process but also ensured consistency across different models.
As you can see in the image below, the consistency is maintained between Trisha and Tonya.
In the final works for the rig, we had to pay close attention to the hips, as they have a lot of volume, and there was a risk of the hands ending up inside the pants.
After that, we could play around with the avatar with no issues at all, making her look fabulous and transporting this pipeline to the rest of avatars to be done.
In conclusion, the final result represents a good balance between the native art style of Regulars and the lessons taught by Botero, creating a body that brings this avatar to life anywhere, anytime.
Overall, the process to create this model allowed us to explore a different approach to designing avatars like this and also a fantastic opportunity to delve into the intriguing figure that was Botero.
Passionate about videogames, movies and creatures. Artist by day and superhero at night.