Making of 'Fall'
Fall, my latest project, was the first illustration I did using a combination of 3ds Max and VRay. After almost five years of experience with mental ray it was a real pleasure trying something new.
This was a project I started few years ago, but it never grew to more than an initial sketch and simple scene blocking in 3D. The first idea for the image was just to create an old rusty factory and something mysterious in that tunnel. But as the image developed and grew, the main story was formed and from that point the whole image began to be shaped that story; it was my main source of motivation and inspiration. I know some artists just sketch and do illustrations, but my way is to create a story first and then create the whole environment around that story. It's basically just a flowing creative process and I think that in time every artist finds the best way to develop his ideas.
The purpose of this Making Of is to show you my working process and to explain some steps during image creation; it's not a tutorial and you probably won't learn something specific like in other tutorials. So, if you want to get some more details about a few of the steps then feel free to drop me email and I'll do my best to reply.
Concept and Initial Scene Blocking
After I got my first idea I made one quick sketch that served as a base for things to come. It doesn't matter if you are good or bad at sketching; it's just that it's the fastest way of making a draft version of ideas that you have. This was made in 20 minutes in Photoshop using simple brushes and some base textures. I'd recommended having a painting tablet for this kind of work (Wacom is the best) because it just feels so natural. Sometimes I do a quick sketch on paper and bring that into Photoshop where I add colors and shading to everything, but since painting on a Wacom is very intuitive it is also a great way to start. So this is how the first sketch looked (Fig.01)
The initial sketch served as a starting point for building the scene in 3ds Max. Using some compositional forms from the concept, I built a few basic shapes in 3ds Max and staged the camera. I believe that the most important thing at this stage of an image is to define the main shapes in the composition, because the whole scene is going to be built around those shapes. The camera is another important consideration because the scene is created from the camera's point of view and so you only need to create what the camera sees. If you don't do this part right then it's going to be very time consuming to change the main shapes later on when you get into the detailing. It's also good to place some basic lights and define some materials ... for this image, I made two materials: one metal and one brick. I then assigned those materials to all the objects (Fig.02).
After the 3D scene blocking was finished, it was time to start with the modeling. When I create a scene with lots of details, it can take up to a couple of weeks to complete depending on the time I have. Sometimes, to make things less messy and help me organize everything, I export parts of the blocked scene to separate files and remodel that area over there before merging it back into the main scene. It's also possible to do XRefs and keep everything in separate files, but that's up to you if you want to try it... I only mentioned it here because for some artists it can work better.
As I said after scene blocking it was time to get into the details and the modeling. This time I decided to model everything I had in scene first, then do the textures and materials later. Sometimes it can be a good decision to separate every process (modeling, texturing, lighting), but most of the time that's not the way I do illustrations because some modeling details can end up in dark areas where I don't need them. The thing is that in time you just learn where you need all the details and where to not put your effort. After a few days of work the model was at the following stage (Fig.03).
I normally use the Edit Poly modifier, starting from a box, cylinder or some other object, or I just draw the curve that's going to serve as a profile and then apply the Lathe or Extrude modifiers. The Symmetry tool is one very important modifier in my workflow because you can get so many different variations from your initial model; you can basically reuse the same model a couple of times without spending so much time building a new one... and the viewer won't even notice that the models have the same base object. Here's an example (Fig.04).
When I am done with modeling for the day I usually do one quick render, bring that to Photoshop and do a quick draft paint-over that image to make a plan for the next day. These over-paints look really simple but it gives me a better visual reference of the details that I need to do in some areas. Sometimes I end up doing only 40% of what I sketched, sometimes it's more, it all depends on how much time it will take and if the new details will fit into the final model (Fig.05 & Fig.06).
Once I was satisfied with the modeling stage I went onto the texturing and lighting. It was not a completely finalized model, but I just decided to stop at that point because it was enough to work with for the texturing and lighting stage. It's always possible to model some more shapes if needed and do another pass of modeling, texturing and lighting.
The final model looked like this (Fig.07).
The background wall is a displaced surface and the bricks under the concrete are 3D models. For the bricks I created a multisub material with three different variations and applied that material to the brick model. Using the Material By Element modifier I randomized those three materials over the brick models to get a natural variation. Keep in mind that when you apply Material By Element you need to have those bricks in one mesh because Material By Element randomly changes material IDs on individual objects inside the collapsed mesh.
Texturing and Materials
Nothing special here; I used VRay materials for everything. I usually don't use anything else except these materials because you can do almost everything you need with them, except maybe high quality sub surface scattering for which there is a new SSS2 material in VRay. The way I am usually working during the texturing and material stage is to build one MultiSubobject material with a couple of different materials that will be used in the scene. For this scene theses were some basic metals, like black, red and gray painted metal surfaces, then more reflective metal materials like bronze, glass and some basic concrete (Fig.08).
Fig.08 - Click to enlarge
After I was done defining the base materials I textured the image based on a combination of those materials. For the UV coordinates, 80% of the time I was using Box UV mapping from the UVW Map modifier. I tried to keep the scale of the Box UV mapping the same for all objects I textured so I could have similar texture scales and a more realistic look. The problem with keeping everything in the same scale is that it's hard because it depends on the scale of the object, so before I did box mapping on every object I went to Utilities and used the Reset XForm option. This kept everything in the same scale so I was safe when applying the Box UV Map. Using this workflow, with multisub material and Box UV mapping, it's very easy to combine materials on the surface. You can make a couple of different variations in 30 minutes and decide which one looks better for you (Fig.09).
Of course base materials were going to be used for some objects in the scene, but not for all of them. The idea here was to quickly texture and combine everything and after that was done, take the objects that were the most important and most visible and redo the UVs and textures on them. So, for example, the front right tower was the one object of that kind. I did a second UV channel on that object and in the Unwrap UVW modifier I used Flatten Unwrapping so that the UVs didn't overlap. Flatten Mapping is separates whole objects into different UV sections based on the angle threshold and it makes the layout of those UVs so there won't be overlapping areas and you won't have distortions. Although this is not very useable for direct painting in Photoshop, it's good for baking dirt. That's the reason why I was keeping this UV in the second channel since I was only using it for the dirt map.
After that I created a material setup for baking the dirt into the edges of that object. I baked that dirt to a separate folder and used that image to drive the Mix map, combining black paint with rusty parts to give the paint a weathered effect. If you don't know how to bake textures, it's in Rendering > Render To Texture. I baked out 4k maps to get as much details as I could. The other reason for that is that since I was using Flatten Mapping, a lot of the UV space was unused, so I was compensating that with a high resolution texture. But if you can spend a little bit more time and create better UVs that use 90% of the UV space then you can probably go with a lower resolution texture for baking.
The material setup for baking and texturing the tower can be seen in Fig.10 - 12.
Fig.10 - Click to enlarge
The lighting was done with VRay's Sun light and one VRay light inside the reactor. It's a very simple setup, and I tried several light setups during the whole texturing/lighting phase. I used several light blockers to get the right shadow placement over the scene. It was done using simple polygon editing to get the shapes I wanted. When positioning light blockers it's good to switch to Light View so you can be more precise - select the light and click Shift+4 for that. You can then get the feeling where your shadows will be cast. The other way to do it is to switch to Orthographic View, match the angle of daylight and then place shadow blockers.
While I am modeling and texturing I like to develop lighting as I go, so these are some of the images I rendered during that process. Lighting is highly dependent on material so that was a big reason why in the end I switched to a completely different light position (Fig.13 - 17).
For the rendering I used VRay, and the DMC sampler. I also used Light Mapping for the secondary GI bounces and Irradiance Mapping for the primary. The DMC sampler tends to behave better in situations where you have small details in models and textures and a lot of glossy reflections so that's the reason I used it. The GI Irradiance map was set to the high quality preset and the Light map to 5000 samples (default 1000). That was more than enough GI to get into every part of the scene.
The DMC sample setup was as follows:
- MIN samples = 1
- Max samples = 2
- Adaptive Amount = 0.95
- Noise threshold = 0.01.
I also used the VRaySincFilter with the size set to 1.5 (default value). Since I was using a Linear workflow from the start, in the Color Mapping section I set Linear multiply, Gamma to 2.2 and turned on Don't Affect Colors (adaptation only). What this did was decide where the DMC sampler used more samples based on the gamma corrected image value, but at the same time it kept the RAW image in linear space and saved it in that way.
I saved everything to Open EXR file format so I could get ZBuffer (included by default) information which I was able to used for all kinds of effects and color corrections in the post-processing phase.
The final rendering was done in 5k resolution, but all the previous test renderings were done in 2k resolution. For the 2k renders it took VRay around one to two hours, which is pretty fast when you look at the complexity of the scene.
Open EXR provided me with floating point values and ZBuffer information in one file. After I imported the file to Fusion the first thing I did was apply the color corrector with Gamma 2.2. This gave me a gamma corrected image that had nice exposure on the screen. After that I used ZDepth to add some FOG, followed by one color corrector and one color gain to make color corrections to the initial image.
If you want to keep a linear workflow all the way to the end, you can leave the gamma setting for the image at 1 (no color corrector with 2.2) and use the LUT table in Fusion to do on the fly gamma correction for the screen. You will find the LUT button on the screen above the flow area. Click on it to turn it on then click on arrow next to it and click edit, type in 2.2 in gamma and that's it. Then, just before saving the file, you can put one color corrector with Gamma 2.2 applied and it will save image with good exposure when viewed on screen. Remember that the LUT table is just to correct it before it goes to the screen in Fusion; it's not going to bake gamma to the image. Using the LUT table is technically the right way to do it and I suppose I will practice this more and more in the future.
Post-processing is always a crucial part of my image creating process and it should be practiced on almost every rendering. Why would you spend 10 hours trying to get same effect in 3dsMax when you can do quick tweaks in Fusion or some other post-processing software? It's like trying to fake global illumination with custom lights when you can use Irradiance mapping and save a couple of days of work (Fig.18).
Fig.18 - Click to enlarge
And here's the final image (Fig.19).
Fig.19 - Click to enlarge
After I was done with the image, it was time to promote it on internet. What's the reason of doing something if you don't share it with other people? There are a lot of great galleries on the internet where I usually send my latest illustrations. Sometimes I get useful critiques and ideas that make me think a little bit about what I did and how it can be better. When working on something personal for more than couple of weeks, people tend to lose an objective perspective of their work. I know I do. So it really helps to see how other people will react to the image. I just want to say thanks for every comment someone gives; it doesn't matter if it's positive or negative, it helps.
To see more by Toni Bratincevic, check out Digital Art Masters: Volume 8
Digital Art Masters: Volume 9
and Prime - The Definitive Digital Art Collection