My experience in 3D stems from a game modding hobby that I have had since middle school. Later on, my interest in modding games slowly transitioned into game development and 3D modeling. I am self-taught and gained my knowledge by following 3D courses and online tutorials. Now I am learning from my current university.
As part of our end-of-semester exams, I modeled a city scene set in Kyoto, Japan. Now let’s get right into it!
Motivation and Reference
We were given a handful of city locations. We were limited as to what we could do or use for tools. We could only use 3ds Max and Photoshop. Software like Substance Painter or any other 3D modeling software was restricted. We could use our own photographs or get textures from the internet, however, we couldn’t use a texture or photograph without editing it. Most of my texture sources are modified textures and photographs from the site textures.com. We were also told to use classic rendering methods like diff-spec and were told not to use things like normal maps.
One of the city locations that were available to us was Kyoto. I find Japan to be a very beautiful country. Since I have never modeled an oriental city before, I found doing so to be very curious and exciting. Looking through photos of Kyoto only reinforced my curiosity and motivated me into settling for a project based on this city.
I then took a virtual vacation to Kyoto through Google street view to gather references for buildings, street objects and just about anything that would help me create the right feeling of being in a street in Kyoto. I was screenshotting, jotting and sketching down any ideas I could have for my diorama’s layout. Apart from Google street view, I also used regular search engine image results for references. I also tried to find any blueprints or architectural schematics to correctly represent the size and proportions of certain elements in my scene. In this case I used a schematic with road width, generic room height and lamp post blueprints (for height reference).
I then placed key references into one image file to create a mood board that I needed to follow to correctly represent the feeling in my scene. I also wrote down any backstory of my scene that I had in mind that I would use for Sketchfab’s annotations.
I created a bunch of grid textures with different scales on them to use during my blockout process in 3ds Max. I then applied the textures to different materials with real-world scale enabled. Now I could start blocking out my houses, roads, and terrain.
This helped me quickly model my environment with proper scaling and proportions and it helped me visually guide myself if I wanted to improvise while working on different elements of my scene.
Heights of rooms in buildings followed schematic references and the horizontal sizing of the rooms was developed the idea in mind that a human should be able to traverse the interior.
Now the main goal was to iterate on the scene as much as possible and save any major changes in a different layer for comparison. I then created some terrain textures I could apply to my scene for visual guidance and to help me imagine how my scene would turn out in the end.
Modeling and Texturing
Now that my blockout was finished, I created a to-do list of objects I needed to model. My process of modeling an object consisted of basic poly-modeling, UV unwrapping and then texturing. I tried to use UE4 naming conventions for better organization and to find a certain asset easily in a massive pile of assets.
Texturing was done in Photoshop and I tried to be as non-destructive as possible in order to be able to quickly change certain elements of my texture if I didn’t like them.
Textures were saved using the Save for Web exporter in order to generate a png-8 file with indexed-color; I did this to have a smaller file size. This approach has its caveats (like artefacts, loss of color) but since I was planning to stream content through a browser I thought that this was a must.
To save time, I created some models that could be reused as decorative pieces for my buildings.
Once the main buildings were done, I integrated them into my main scene file using XRef. I then did touch-ups to my terrain while working on other props for my scene.
Once all the scene props had been modeled and placed, I started working on tiny details and added finishing touches to the scene like telephone polls, leaves, rocks, and pretty much anything to liven up the scene. I used object paint to place these tiny details, not by hand but by using a brush.
As a final touch, I added trees to my scene. After modeling them using regular poly-modeling, I vertex painted the branches to add leaves and used a particle system to add the branches to the tree.
Getting Ready for Upload
Once my entire scene was done, I created a second UV channel for my houses and baked an AO map for the walls to boost contact shadows on the buildings. I had to move ground decals higher than usual to avoid clipping and fixed up naming conventions, deleting unused materials and removing any guide objects that didn’t have to be uploaded.
The first thing I did was set rendering to Classic since that’s what my scene was aiming for.
I gave my scene an environment map, which would give me lighting and mood similar to my references, and added subtle post-process effects to enhance my scene.
And of course, as a finishing touch, I added annotations with my scene’s backstory!
If I found a mistake or had to fix something quickly, I updated my scene in 3ds Max and reuploaded it. What I like about Sketchfab is how user-friendly the 3D editor is and how easy it is to fix scenes if you have to reupload.