Turning Summer Bike Rides into Peaceful 3D Environment
Dmitry Kremiansky shared the work process behind the End of Summer project, showing how his bike rides transformed into the detailed scene and explaining how the water and vegetation were set up to paint a realistic picture.
Introduction
Hello! My name is Dmitry Kremiansky, I am a Senior/Lead Freelance 3D Generalist based in the UK with over a decade of experience working in VFX, Animation, and VR.
I've always been passionate about drawing and painting, and I was making models out of scraps as far back as age 6. But PC gaming ultimately drove me to start my journey in 3D back in 2013, first using ZBrush to make characters inspired by Assassin's Creed 3, then slowly pivoting to props and environments in Maya.
I've had many great mentors over the years, from whom I've absorbed information like a sponge, and through a mix of impostor syndrome, ambition, work ethic, and lots of hard work, I was able to prove myself to my harshest critic – myself.
Since I've spent half of my career working in cinematics on very tight schedules, I was able to accrue over 50 professional projects during those years, not to mention dozens of personal ones. Some of my favorite recent projects include Shōgun, Disenchanted, and my workshop for The Gnomon Workshop.
End of Summer
I really like classical paintings of pastoral scenes, and after cycling many times along the waterways around Glasgow, I felt inspired to create something of my own. I also wanted to experiment with the feasibility of using Unreal Engine 5 Lumen/path tracing for my personal projects and to add more movement to my scenes.
I ended up taking lots of videos and photos during my rides, as well as shooting a bunch of textures and scans. I had some ideas for the composition, but I also wanted to keep the project manageable and avoid spending too much time on it, so I didn’t aim for a specific concept and instead followed the path of least resistance in places.
The project sat collecting dust on my drive for years, but over the holidays, something clicked in my head. Once I reopened it and migrated it to Unreal Engine, it was done very quickly.
Composition
I firmly believe in the rule of thirds for composition, and over the years, I developed some preferences and an instinct for what feels right and wrong.
I usually first add a ground plane, then set up a camera using settings and focal length I’d use in real life for this type of scene, and quickly block it out using a library of 3D assets I made in my spare time over the years, some of which are available on my ArtStation, CGTrader, and CGLounge.
After, I add boxes and blobs where I have missing assets, lock the camera in position I like, tweak the placement of the assets to create clusters of objects against plainer areas to let the eye rest, and then more refinement, more textures, more everything.
The church was made using box modeling, followed by beveling, UVs, and texturing in Substance 3D Painter. This particular asset uses UDIMs, so it isn’t intended for games, but rather for VFX. However, for my specific use case, I found that it wasn’t slowing the scene down too much.
The water is a plane with different textures panning across it, using normals for the waves and a mask to drive the foam and bubbles on top. I also added a couple of planes with algae textures and alpha to achieve parallax and create the illusion of volume.
During my trips, I noticed that canal water is not as clean as it’s usually made in CG, but instead has lots of detail – from oils and foam to rotting plants and mud – so it was important to me to achieve that look.
For the vegetation, I used PaintFX as a starting point for the nettles, grass, water lilies, and cattails, while the trees were made in SpeedTree. I saved a lot of time using the church asset I already had for a while, and I could've used Megascans or other libraries for grass to save time, but since I wanted to have assets I own and can reuse, I ended up making those myself.
I think on personal projects like this, it's important to find a balance between saving time to be able to finish the piece, enjoying the process, and creating stuff you can reuse, because when you work in the industry professionally, you generally don't have the luxury of time to experiment or try more than one approach to deliver an asset.
Texturing
There was no need for retopologizing for this project, but generally, I use Maya or ZBrush: the more organic the object is, the more you can get away with just using ZRemesher, and I always use Maya for unwrapping. These are the UVs of the church; the time I put into laying them paid off tenfold during texturing.
I use Substance 3D Painter for my texturing work. It’s an amazing tool that I’ve been using on a daily basis for almost a decade, and it's amazing how far it has come. I have a 7-hour-long workshop with awesome folks from Gnomon explaining in detail how I texture, but in a nutshell, I generally create a few layers using tileables/triplanar textures and cover as much of the surface as I can before adding detail, whilst making sure the bricks wrap correctly around the corners and align with most of the detail, like so:
I then proceed to add details using a few fill layers with masks in a folder to introduce breakup, which I can then duplicate across different sets, and make sure that bespoke parts such as the arches over the windows get their own treatment so the texture works well with the model.
I then finish duplicating and adjusting the detail on top:
Voilà!
For my asset library, I generally create assets with a medium amount of age detail, as I feel it wouldn’t make sense to have all the dirt and damage we see today on a 150-year-old church if the setting of the project is when it was newly built. Then, when needed, I simply add dirt procedurally on top and darken the details to age the asset.
Final Scene
I worked checking through the camera constantly to make sure I was not spending time adding dressing outside the field of view. Once I set up PCG for the grass scatter, it was very easy to apply the same setup procedurally to multiple ground meshes, and then it was simply a matter of tastefully adding detail where it was needed: moving trees around, adding hales of bay, birds, etc.
I also added a couple of alpha planes to introduce better layering with fog and dust between elements. To have some more interesting lighting on the hills, I added some giant blobs made of spheres out of frame as clouds and moved them around in the sky until it looked good.
In the end, I added some falling leaves and moving dust for extra oomph.
Lighting
I used a directional light to bring out some details and an HDRI and used Lumen with pretty much default settings. In the PP Volume, there is a tiny bit of chromatic aberration added, as well as some subtle grading lifting the shadows, a LUT file, and some other magic.
Conclusion
Over the years, I've started a lot of personal projects, all of which were things I'm passionate about, but only the ones where I managed to cut the scope to something sustainable and didn't get lost in the process until I got bored were finished.
Time is your most valuable resource; therefore, when working on a large personal project, find ways to continuously reduce scope whilst improving quality by focusing where it matters. Reuse small props and textures across the scene, set up a camera angle/path early to focus on things you will actually see, and block out large chunks of the image first. Do not use hundreds of 8K textures for a simple asset; tweak assets separately to save render time. Have a plan for the next step so you know where and where not to cut corners.
