Hello again, my name is Rafael Diniz, and I'm from Rio de Janeiro, Brazil. I started in 3D in September 2018 with the purpose of complementing what I was doing in Photoshop,
but it turned out to be a passion, and the main tool I’ve used since then has been Blender.
The work I pick up is usually related to advertising. I've worked in a studio on post-production, which opened my mind and made me want to use 3D more than Photoshop. Of course, it was not an easy start, following those YouTube tutorials, then always researching more, testing things, courses, plugins, everything that could improve my workflow, but logically always trying to understand the technical part, always wanting to learn why something happens, to learn to control.
This time I was trying to create something that had smoke. I put that idea in my head and went after references. I ended up getting to know the Formula Drift, which ended up being my biggest reference for the project.
Again using Blender as the main tool, this time I decided to test the new features, such as OpenVDB and Adaptive sampling. For the smoke, I risked using Houdini for the first time. The trees were made using SpeedTree, some textures from Quixel were used, and the track marks were made in Substance Painter. The video post was made in After Effects, the sound I edited in Sony Vegas, and the cover post I edited in Photoshop. I believe that I have never used so much software on one project before, haha.
I believe that this was the best elaborated and most well-optimized project that I have done. Because I knew it was going to be an animation in Cycles, I already knew it would be heavy, and the focus was on being as realistic as possible. From that I needed to put together a plan of how this would work; the first thing was to create the separate files and then just link them. Most of the props were created in a separate file and then linked to the environment file that was later linked to the final file—the memory usage at peak render was a little more than 3GB. Another thing that made a difference was Adaptive Sampling from Blender 2.83—I believe the rendering time has decreased by ~20-30% using it.
Another thing that I had never tested before was using planes like projections. The mountains, bleachers, people, etc all created using this technique, and I confess that I was surprised—with only a few color gradient adjustments in the shader editor, things fell into place.
People who follow my work know that modeling is not my focus, but I will model simple things ^^. I downloaded the car in this scene, for example, but can no longer find the link to share it with you. Most of the props in the scene I modeled, and the scene also contains a few props from Quixel.
Now comes my favorite part. Let's start with the car, where my base came from the course “Master Car Creation in Blender”, but this time with a white car history changes a little. This was my second project with a white car (the first project I didn't post but I intend to next month because there are many images), and it's hard to really compare it to the others—you will notice that the nodes are a mess, haha.
The texture of the track is from Quixel and the tire tracks I created in Substance Painter. To mix the two, I used only the Diffuse in multiply mode.
Simulation and Render
Of course, my first option for creating the smoke was Blender, but after some tests, I saw that it would be difficult to achieve the expected result. I already knew that there was OpenVDB in the new versions of Blender, so I settled on Houdini. I took a few days to study the software first and it amazed me a little in the beginning. After a while, I discovered that although I loved the software, it doesn't get any easier the more you learn, because you want to test more difficult things and sometimes it seems impossible. The control Houdini gives you is incredible. There were no web tutorials of what I needed, so I had to really improvise and test. Although this part of the project was the one that I had planned out the most, because it was something new, this is where the most problems appeared. The simulation took 2 hours and 48 minutes.
As part of my workflow, I like to separate the passes that I normally use in compositing. This is also when I fixed other things like smoke. Among the many other things that I did in post, I used the Volume Direct pass to improve density, and the Gloss Indirect and Direct passes to ''jump'' the reflexes more. The rendering time took about 35 hours, 600 frames at 120 frames per second.
- CPU: Ryzen 5 2600
- GPU:Gtx 1070 8GB
- RAM: 16GB
About the Author