I have read all the comments and suggestions posted by the visitors for this article are very fine,We will wait for your next article so only.Thanks! order now
Steffen Hampel recreated a cinematic photo by Alex O’ Dowd in 3D using online resources and Megascans and did a breakdown of the scene.
Hello, I’m Steffen Hampel. I am 25 years old, born and raised in northern Germany. I love playing video games and I’m also very interested in eSports.
When I was 13, I got in touch with VFX & 3D by randomly watching breakdowns on YouTube. I got interested and since then I basically watched every single VFX breakdown on YouTube to learn how things work. In 2017, I learned about the possibility to study VFX & 3D animation and I instantly knew I had to do it since that it was what I wanted to do my whole life.
After the first three months at the university, we started learning the basics of 3D and I got fascinated right away. It was just the fact that you could create anything and when done right, you cannot really tell if something is simulated or real.
In April 2019, I finished my diploma in VFX & 3D animations. So, I am just a recent graduate hence I have not really worked on any professional projects besides a smaller environment modeling job for a Chinese feature film.
Inspiration & Reference
My main inspiration is definitely the photography of Alex O’ Dowd. I randomly stumbled upon Alex’ cinematic photography and I just fell in love with his pictures. I love Japanese architecture and the composition was so interesting that I knew immediately that I wanted to recreate his photo in 3D. I always wanted to recreate an image in 3D anyways and this was a perfect chance.
In the beginning, I wanted to recreate the image 1:1 but after some time I learned that was pretty much impossible. The first thing which came to my mind was the trees. What species of trees are these? How shall I recreate them in SpeedTree? So I just tried to go as close as possible within my skill range.
I contacted Alex and he actually helped me a lot by sending me the original image, so I was able to get the metadata of the photo. Adjusting the camera settings in Maya to be the same as in the photo was the first step. After that, I had to build a basic blockout and try to fit the camera position. Since the building had pretty clear lines I was able to find the right camera position pretty fast. Then, I added some smaller objects and matched them with the photo. Luckily, Alex also sent me a google street view link of the real place. That really helped me figure out where objects stand in 3D space, the light sources, etc.
I modeled everything but the big rock in the foreground (which is a Megascans asset) and the scooters in the garage (found for free on the internet). Everything was made in Maya. The key object was the actual police station. First, it was really hard for me to figure out the structure of the building itself. Even with the google street view, it was kind of weird and difficult to understand how it was made. After analyzing multiple reference screenshots I was able to build the mesh. It is not perfect but luckily I could get away with it due to the camera angle hiding a lot of things!
Next, I started modeling smaller props like the windows, roof tiles, and pipes. They are all very simple geometry, nothing special. When I modeled the air conditioners though, I used a photo from cgtextures.com as the main reference. I placed the image in Maya and since it was a frontal photo it was very simple to just rebuild everything I could see. I duplicated the final model and changed minor details so it did not look like I just duplicated the air conditioner. For the cables, I used a python script which is able to convert curves to a cable geometry. That saved a lot of time. I created the trees in SpeedTree starting with the sample broadleaf and then adjusting a couple of things. Other trees which I was not able to recreate, were just bought on cgaxis.com. At this point I noticed that I really have to get better at SpeedTree.
Creating materials is a fun part for me. Especially since Quixel Mixer makes so many things very simple and effective. I use Mixer mostly for things like grounds, streets, etc. and Substance Painter for special objects only (when I need to create a really special look). Sometimes I also go old school and use Photoshop. I like to use base images from cgtextures.com and play around with them, make them tileable and seamless. Depending on the scene I create a glossiness map and adjust it in lookdev later (I use the same glossiness map as a bump map). You can also do a lot with procedural shaders in V-Ray.
For this scene, the wall is actually a shader with a dirt material network. I did not layout UVs, instead, I used the V-Ray triplanar node. I connected a V-Ray dirt material to the diffuse channel and a nice grunge map on the radius to create dirt with the ambient occlusion pass. The nice thing here was that with this method, I would get only dirt where things were placed, for example, only on window frames, electric boxes, etc.
In general, I prefer using tileable textures for scenes like that since it saves A LOT of time. Especially when you create multiple shaders with the same textures but use them differently. I used Substance Painter only three times for this project. Firstly, on the stonewall in the foreground, secondly, for the big concrete pillar in the foreground and thirdly, for the red metal pillars. I baked ambient occlusion, world position pass, and curvature. I used those passes to create things like dirty/worn edges. I then used different brushes to create different layers of dirt trying to match the textures with the photo.
Concerning the Mixer/Megascans combo part, I really like to create streets with fresh asphalt and then add some imperfections in the Diffuse and Displacement channel. For the ground, I used scans like “Mud Tracks 2×2 M“, “Asphalt Rubble“, “Dried Leaves“, “Sand 2×2 M” and “Soil Mud“.
Another special part for me was recreating the glass material with the Japanese sign and the signs in the foreground. I actually was able to figure out what the signs say in the original image so I could recreate the exact same letters and color. Sadly, I did not know the exact same font. The three signs on the left side of the image are also actual maps of the real place. Small details like that definitely helped me selling realism in the image.
What Megascans Materials to Choose
It really depends on your scene. If I had to choose though, I would definitely recommend every brick material and street imperfections like asphalt patches and potholes since they are really detailed. Also, things like rocks and nature stuff from them are really insane. For another scene, I also used a lot of their Icelandic scans.
The lighting in this scene was definitely the hardest thing to do. For me, it was really tough to place all the lights to make them look natural. I really had to analyze the original image quite a few times to understand where all the light sources are and how I could achieve correct shadows. I started with a free night skydome from hdrihaven.com. After that, I placed a couple of rectangle lights with a fluorescent lamp HDRI. I always used directional lights with V-Ray values around 0,8-0,9. When I had the basic light setup done I created the four important lights: the inside of the police station, the lamp in the park on the left side, the lamp on the right side in the background, and the green lights above the garage. It was all a game of adjusting intensities and directions. For instance, the inside of the police had lights on the ceiling and one big directional light above the entrance which was aiming at the street. I constantly changed the settings of lights because I always compared them to the original image and really tried to match them.
When working on scenes like this, I recommend analyzing in depth how light behaves in original photos. Shadows and reflections often help a lot to understand where the light sources are and how strong they are.
I definitely learned a lot about lighting and how light behaves. Also, I learned about really taking care of small details like reflections in metal or shadows to conclude where light sources are. Another point is that I realized how important it is to always stay close to real-world reference in order to achieve photorealistic results. My picture is not perfect of course, but I bet next time I will be able to get closer to photorealism. It’s just a matter of exercise. Furthermore, I learned how to get away with murder in some parts! Sometimes, it is just more efficient to work dirty and fast on those parts that will not be seen to save time as long as the result works.
Steffen Hampel, 3D Artist
Interview conducted by Kirill Tokarev
Some pine lookin’ Bark collection from Quixel: