logo80lv
Articlesclick_arrow
Research
Talentsclick_arrow
Events
Workshops
Aboutclick_arrow

Okino's Residence: Creating a Studio Ghibli-Inspired 3D House

Ana Cristina Pacheco shared the workflow behind her project, Okino Residence, discussing the modeling and texturing workflows she used to achieve an environment with a Studio Ghibli style.

Introduction

My name is Ana Cristina Pacheco, and I recently completed my MA in Games Art and Design at the University of Hertfordshire. For my final project, I recreated the Okino Residence, a Studio Ghibli-inspired house, as a full 3D environment.

I first started experimenting with 3D as a hobby, sharing my work on platforms like ArtStation and Instagram. Over time, I became more focused on 3D environments, which led me to pursue a master's in games art and design at the University of Hertfordshire. During my studies, I honed my skills with Unreal Engine 5, SpeedTree, Substance 3D Designer, and other industry-standard tools.

Before creating Okino's Residence, I worked on a variety of smaller projects, some realistic, like a Victorian room modeled in Maya, and others more stylish, like hand-painted characters in ZBrush, one of them a sculpture for Circus Macabre, a video game from the University of Hertfordshire. And more projects that I hope to share with you soon.

Concept & References

At the start of the project, I explored several possible directions before deciding on Studio Ghibli as my primary influence. I made my top 8 of concept art that could work to enhance my CV, as I was looking for concept art that is:

  • A Complete environment.
  • An outside environment.
  • With lots of vegetation to make in SpeedTree.

After sending the top 8 concept art that I found that matched what I wanted to do, I sent them to 5 distinct groups of friends and family to see which one they liked the most. And, all the groups confirm my choices as Studio Ghibli art was the favorite option, or in second place, in all 5 groups.

Finding references after this was easy. I gathered references from both Ghibli films and real-world European houses, as the architecture in the original animation draws heavily from European influences. Genshin Impact, as well as Animal Crossing, were other important references. I wanted references for how to do stylized plants, but in 3D.

Blockout & Layout

For the blockout, I started with very basic geometric shapes in Unreal Engine to define the overall proportions of the house and garden. The blockout stage focused on proportions, silhouette, and overall composition. It also allowed me to set up the main camera angles, since I knew the presentation would be a big part of the final delivery.

At this stage, I also set up a simple Directional Light and sky to test shadows and lighting direction. Sharing this early layout online with my professors was useful. I received feedback on adjusting my camera angle. After adjusting, I gathered more feedback that confirmed the proportions were working well, which gave me confidence to move into the next steps.

House

For the building, I used Blender for the base modeling, keeping everything modular. I planned to redo many of these assets later. For example, I expected to rebuild all the modular 3D meshes of the house that I did in Houdini for efficiency and modularity. However, when I shared the Unreal version, I was surprised by the positive reception to the house model and its textures.

This feedback changed my priorities: instead of remaking the house, I invested more time into the foliage, as it was too noisy. This decision saved production time and allowed me to concentrate on the element that would ultimately define the project's identity. It also highlighted the importance of external feedback, as it made me make smarter choices, and such saved a lot of time.

Foliage: The Core of the Project

Foliage became one of the biggest focuses of the Okino Residence. From the beginning, I wanted to push myself in vegetation workflows, and SpeedTree became my main tool. My initial approach leaned toward realism, following common tutorials, just to grasp the software better.  When I put them all together, I already knew I needed to change them, but this gives me the perspective of what to change.

The main problems were that plants looked noisy and lacked the clarity needed for readability at a distance. To solve this, I studied games with stylized foliage, especially Genshin Impact. I broke down how their foliage achieved visual simplicity without losing richness:

  • Separation & Rule of Three: Plants often appear in small clusters, not overwhelming chunks, making them easier to read.
  • Rotation: Make flowers in different vertical positions so that when they rotate, they can look like a new plant.
  • Exaggeration: Larger petals and leaves improve silhouette readability.

Applying these principles transformed my foliage. Flowers like Loosestrife, which previously looked flat and stick-like, became fuller, easier to read, and more stylized. By exaggerating their forms and carefully spacing clusters, they fit into the Ghibli-inspired style more naturally.

After I figured out the style, I made two versions of the flowers, for when they are further away and for close-ups.

ZBrush & Kiki

While much of the work was procedural, I also sculpted details in ZBrush, particularly for flowers and leaves. Baking these high-poly sculpts into low-poly meshes helps create smart materials. Another extra detail included a small model of Kiki (from Kiki's Delivery Service) as a scale reference in the environment. This helped ensure proportions stayed believable while keeping the Ghibli style.

For this project, I made sure that all my assets were modular. So they could be reused efficiently and support the creation of smart materials for procedural workflows. This approach allowed me to save time and quickly produce multiple iterations. The only exception was Kiki, which I created as a small extra once the rest of the environment was complete.

Materials & Procedural Texturing

Since I already made the ZBrush HP, I decided to make the textures for the plants in Substance 3D Painter. I combined procedural, basic smart material bases with hand-painted detail, which aligned with the project's stylized approach. And after that, I joined the textures in Atlas, and with that set up, the foliage was ready for Unreal.

The last and most important thing for the foliage here was to create the color variation shaders in Unreal. I did it by changing the SpeedTree master, which has its pros and its cons.

  • Pros: Immediately after you put in foliage, all your color variations are set, and you get to use the wind from SpeedTree.
  • Cons: Since it was a plugin and not saved in the project, but in the program. I could not render my project on the university's computer, and at one time, all my shader changes were erased, so I wasted one day of the project putting it back, even though I had copies of the project.

What I implemented was:

  • Color variation by instance.
  • Color tint, with a color mask for changing the color of the flowers.
  • Noise texture so the same foliage has variation.
  • And to help sell the Ghibli style, the 4 points are changing the emissive color to give a little bit more flatness to the foliage.

House, Ground, and Bark

Procedural texturing has become one of my strongest technical areas. Using Substance 3D Designer, I created tiling materials for grass, plaster, brick, roof, etc. What I enjoyed most was the control these materials gave me. For example, my grass texture could be adjusted procedurally to show more or less dirt, or to include dandelions in varying amounts. This flexibility helped maintain consistency across the environment while allowing for quick iteration.

Overall, my workflow in Substance 3D Designer was this for almost all the materials:

  • The blue color is the height.

Each blue square in the Height Map is a different type of tile. And I joined them all to create variations to the roof tiles, so they do not all look the same.

  • The green color is my brushstroke set.

For the brushstrokes, I make the same as with the tiles, but to have a bunch of color variance and some deformations to the original shape, this brushstroke set I use in all my other textures to keep a consistent look.

  • The white one is the flood.
  • The black one is a random grayscale.
  • The black and white are more to add color variation and height variation to an entire roof tile.
  • And orange is the base color.

Unreal Engine & Shader for Material Layers

In Unreal Engine 5, I experimented with its newest features. I used Nanite for tiling materials to maximize performance without losing detail. Splines created the stone pathway, ensuring smooth placement and modularity. Runtime Virtual Textures allowed the grass and ground to blend naturally with surrounding elements.

The achievement I am most proud of was building a triplanar, layered texture material for adding the tillable textures from Substance 3D Designer. This setup combined Nanite, triplanar projection, and color layering, giving me unprecedented control. From my knowledge, this workflow only became possible after Unreal Engine 5, making it a cutting-edge solution. It gave me confidence to build shaders on my own, as it was such new technology that it didn't fully combine every single technique in one video.

My process was this: For the ML, I use the triplanar setup from Visual Tech Art and put all the information in a Material Layer. In the ML, I blend the two Material Layers, adding a height lerp for the moss to grow on top of the roof or from the bricks between the seams.

Tips on Assembling the Environment

Assembling the final scene was easier because, from the start, I placed all assets in the environment. This made gathering feedback simple and showed me where more attention was needed. It also kept me from getting stuck on a single asset, helping me focus on the project as a whole. Working this way lets me apply feedback quickly and manage my perfectionist side. Details like leaves, smoke, butterflies, and even Kiki were added at the end, and they enhance the scene, but the priority should always be the main environment, so time isn't lost on less important elements.

Lighting & Rendering

I had the main Directional Light set from the get-go, but finding the correct lighting and color for the postprocess required a lot of feedback. What I recommend if you have a reference, check the values, so they are kind of matching the reference, which helps a lot.

The picture on the left shows the reference, while the one on the bottom right is from the middle of my post-processing adjustments. There were still some things to refine at that stage. What really helped me was constantly taking screenshots of the project and sending them to Photoshop to check how the values were working.

After that, I did a final pass in DaVinci Resolve, mainly adjusting contrast. I didn't want to change too much, since my goal was to have the color and lighting already set up properly in Unreal Engine.

Conclusion

Some of the biggest challenges included:

  • Stylization Decisions: Finding the right level of simplification without losing detail.
  • Optimization: Ensuring assets are as modular and procedural as possible, keeping the hand-painting feel.
  • Iteration Pace: Accepting that redoing everything isn't always the best path; sometimes feedback tells you what areas need more focus.

Overall, this project helped me a lot to grow as an artist in Unreal Engine 5, SD, and SpeedTree. I want to thank my professors, Neil Gallagher, Wayne Robson, and Eden Anthony, for their guidance, as well as industry professionals Caroline Couture, Guillaume Hecht, Zachary Allen, and Santiago Deus for their valuable insights. I am equally grateful to the online communities on Polycount and Discord, the Discord communities: Stylized Station, Tarmunds' Tavern, and ZugZug Studio, whose feedback pushed me to refine and improve.

I think the biggest lesson from this project was the importance of asking for feedback and planning time for it in the Gantt chart from the start. Reaching out on ArtStation to artists I admire and asking specific questions about the details I wanted to improve made a huge difference. Being open and not shy about seeking advice truly helped the project grow in the best way.

Ana Cristina Pacheco, Environment 3D Artist

Interview conducted by Gloria Levine

Join discussion

Comments 0

    You might also like

    We need your consent

    We use cookies on this website to make your browsing experience better. By using the site you agree to our use of cookies.Learn more