Combining PBR & Hand-Painted Texturing in Stylized New York Cafe Scene
Marina Melo talked to us about the NY Cafe project, discussing translating a stylized 2D concept into 3D, from modeling in Maya to a texturing approach that blends PBR workflows with hand-painted techniques using Substance 3D Painter.
Introduction
My name is Marina, and I am a Brazilian 3D artist mainly focused on environments and props. I enjoy working with stylized art, but I am also interested in exploring realism.
I have always loved art, something that has been part of my life since childhood. I started drawing traditionally and later transitioned to digital art. At some point, I discovered 3D modeling, and it immediately sparked my curiosity. At first, everything felt quite intimidating, but the more I studied, the more I realized I had found something I was truly passionate about.
I have been studying and practicing 3D consistently for about a year. I started using Blender, mainly through tutorials and by exploring the software on my own. After some time, I began taking online courses and eventually moved to Autodesk Maya, where I was able to deepen my understanding of modeling fundamentals.
Today, my main workflow is polymodeling in Maya, although I also enjoy exploring digital sculpting using ZBrush and occasionally Blender as well.
Recently, I participated in the development of a small game in Unity called Folia Rush. In this project, I was responsible for modeling low-poly assets and creating the textures. It was a very enriching experience, as I had the opportunity to work within a game-oriented pipeline and better understand how assets integrate inside a real-time engine. In addition, I continue looking for new opportunities and experiences in the field to keep improving as an artist and expand my portfolio.
Inspiration & References
The idea for this project came up while I was browsing Pinterest, looking for references to create a new stylized environment. That’s when I came across a concept by the artist Igor Rozovny that immediately caught my attention.
I liked the visual language of the drawing, especially the way he works with shapes in an expressive and slightly exaggerated way. This motivated me to try recreating that aesthetic in a three-dimensional environment. During the modeling process, I tried to preserve this visual language so that the final result would maintain the same spirit as the original artwork.
For references, I mainly used the concept as a guide for proportions, composition, and the overall silhouette of the scene. For the materials and surfaces, however, I looked for real-life references to understand color variation and surface texture better.
Composition, Blockout, & Modeling
The first step was creating a simple blockout, using basic shapes only to establish the overall composition of the scene.
From the beginning, I also set up a fixed camera in the final render position, since the concept has a very strong perspective. This helped me adjust the objects while already considering how they would appear in the final image.
From that initial blockout, I started gradually refining the shapes and volumes of the building and the street elements.
An interesting example was the upper part of the building, where there are several stacked stones. I modeled a base stone and used the Array tool in the transform tab to distribute them along the structure. After that, I adjusted each one individually to make the result feel more organic.
In the lower part of the building, where there are wooden elements, I chose to model only the main beams and leave the plank divisions for the texturing stage. This helped keep the geometry more optimized.
I also used Lattice quite a lot to deform and adjust shapes without compromising the topology, which was useful for reinforcing the more stylized look of the project, especially while working in Autodesk Maya.
Retopology & Unwrapping
Since I mainly work with polymodeling, the assets were already built with the final topology in mind, so there was no need for retopology.
For the UV unwrap, I decided to work using UDIMs, as this gave me more freedom to organize the assets and maintain good texture quality across different parts of the scene.
First, I organized the objects into groups that made sense to share the same UDIM, such as the building, the street, the props, and the windows. Most of the unwrapping was done using RizomUV, which I consider an extremely efficient tool for this type of work. Some simpler objects were unwrapped directly in Autodesk Maya. I also used UV stacking on some elements with repeating shapes to save space and maintain good texture resolution.
Texturing
Since I was texturing several assets in the scene at the same time, I enabled self-occlusion only by mesh name during the baking process to avoid intersection issues or unwanted shadows between objects.
I usually work with a PBR workflow, but for this project, I wanted to mix a bit of PBR with hand-painted techniques to get closer to the aesthetic of the original concept.
One of the first things I do in Substance 3D Painter is replace the default lighting with a more neutral studio HDRI, as this helps me evaluate colors and materials more accurately. I start by creating a Fill Layer with the base color of the material, usually removing the initial roughness, so I can build this channel gradually as the material develops.
This scene has a strong presence of color variation, so I used the default hard brush to paint subtle changes directly into the Base Color. On the upper part of the building, for example, I started with a base red and added a slightly purplish gradient using Multiply. After that, I introduced other subtle color variations such as blue, orange, pink, and yellow. Some of these colors were softened using blur, while in other areas I kept the brushstrokes more visible to reinforce the hand-painted look.
After working on the Base Color of the upper part of the building, I started building the surface material. For this, I used a base material from the Substance library, removing the original Base Color to preserve the colors I had already painted. From there, I mainly adjusted the roughness and height channels and added some edge wear effects using masks, refining the material until I reached the desired result.
This was a more specific example of the process, but I followed the same texturing logic for the other elements in the scene, always starting with the base color, adding hand-painted variations, and then building the other material channels.
Throughout the process, I also used real photographic references of materials to better guide the surface details.
Assembling the Scene
Once the texturing process was finished, I exported the textures and finalized the scene assembly in Autodesk Maya, using Arnold for rendering.
Since the main camera had already been defined from the beginning of the project, I only needed to make small position adjustments to some objects.
The original concept already featured several elements distributed throughout the scene, so I tried to place them in a way that stayed close to the layout of the artwork, making small adjustments in position and rotation so they would work better in a three-dimensional space.
Lighting & Rendering
For the lighting, I mainly used Area Lights, combined with an HDRI to complement the ambient lighting. In total, the scene was lit using five area lights.
I placed two cool lights at the back of the scene to better match the lighting seen in the original concept. In the center, I used a larger area light with a neutral temperature, working as a softer base light to balance the overall values of the scene.
At the front, I added two warmer lights to better highlight the facade and bring more definition to the volumes of the building and the props. This contrast between cooler lights in the background and warmer lights in the front helped add more depth to the image.
The HDRI was mainly used to enhance the sense of natural lighting and help complement the overall illumination of the scene.
The final render was done using Arnold in Autodesk Maya, and in post-production, I only made subtle adjustments, such as noise reduction and small exposure corrections to better balance the final image contrast.
Conclusion
It took me about two weeks to complete this project. One of the main challenges was translating the stylized language of the concept into 3D. In the original art, the exaggerated shapes work very naturally, but when bringing them into three-dimensional space, it was necessary to make several adjustments so the forms would continue to work well.
At certain moments, I had to redo parts of the modeling, adjusting proportions and angles until I found shapes that better conveyed the style of the original artwork. This process ended up being an interesting exercise in interpreting forms, trying to preserve the spirit of the concept while adapting the elements so they would work convincingly in 3D.
One of the main lessons I took from this project was realizing the importance of well-resolved shapes. It is not always necessary to create extremely complex geometry to achieve a strong visual result. Many times, working well with the basic forms is already enough to communicate the idea of the scene.
For artists who are just starting out, my main advice is not to give up halfway through a project. A piece may seem difficult at the beginning, but finishing it always brings learning. In many cases, it is better to complete something and improve in the next project than to get stuck trying to reach a level of perfection that prevents progress.