How to add destruction to a 3D scene part 2: after the flood
Now we move on to the most fun part – destroying the city and replacing the concrete and all the man-made stuff with foliage and nature. It’s a timeless juxtaposition and will never stop being alluring, especially to a concept artist or illustrator! Full disclosure, I had planned to do this series irrespective of the new Last of Us game, and kind of wanted to get it done before I saw the concept art so I couldn’t get too influenced by it.
I hadn’t seen any while working on the ‘before’ shot, and then the art came out right before I started the ‘after’ shot and, well, maybe it did influence me a lot. It was all just too good to avoid massive inspiration. Damn those Naughty Dogs. So let’s get started taking our grimy, bustling city, and playing with what happens when a flood totally destroys it and all the people leave (I do promise that I at least thought of the flood idea before I knew anything about The Last of Us Part II)!
It’s not too easy to find good reference for an overgrown, apocalyptic city. Especially not a flooded one. But for this, I’m okay with winging it and just having a play. I’ve done a few such images before and sometimes, if you tried to go with realism, you might miss out on some creative ways to show nature reclaiming the world. For my image, I definitely wanted water on the ground somehow, ideally forming a kind of stream with some distant waterfalls. Plus, as I said in the intro, I was starting to see the avalanche of The Last of Us Part II concept art and so I gathered as much of that as I could for inspiration.
Other than that, I found certain images of Chernobyl worked quite well, although not too many images of that region served as particularly useful because many of the buildings themselves were not overgrown and that was something I wanted to depict. To that end, I just searched Google for images of buildings that were overgrown with ivy.
Resetting the camera
I struggled with this for a while because I ideally wanted to keep the exact same camera set up so the before and after would appear seamless in that regard. The lens in the first shot was fairly long compared to my typical set up of around 20-25mm, and longer lenses can work well in city shots. For this shot though, I wanted a wider scope to fit in as much natural landscape as possible. You’ll notice I show more of the ground plane in this new shot.
I really need to lead the eye in, and pointing the camera down a little and/or expanding the frame vertically is useful when you want that effect. An upward shot can be dramatic, and is a little on the cliché side of concept shots, but you’d be amazed at how helpful it is to pull that horizon line up and allow the contents of the ground-plane to draw you into the picture.
All along I knew I wanted a big boat to act as the focal point in the scene. I downloaded this from Sketchfab and painted rust all over the textures within Photoshop so it was essentially just rusted all over. I also added some ground debris from Kitbash3d, but really that was just to give me an approximation because much of the ground plane would be replaced. At this stage I was actually under the assumption that I would be using full 3D for the ground plane and its contents.
To help the focal point pop I’m generally adding pockets of volumetric fog here and there. I’ll go into atmospherics in more detail in the next step, but in these test renders I’m also rendering out a ‘mist pass’ in Blender and quickly comping it over the top in Photoshop just to get an idea of how it affects the read and the depth (essentially, a mist pass is like Z-depth).
Atmospherics & volumetrics
I’m a huge proponent of volumetrics and fog. It’s a necessary tool for any cinematographer. Hell, cinematographers even add subtle fog to interiors to help things read. In this scene, you’ll notice I have one large box covering the entire scene with a principled volume shader applied. I also have one volume VBD file in the middle of the scene (the smaller wireframe box in the middle). Both are set to very low density (I’ve increased their density in this shot just so they’re more visible). Importing VBD files is new to 2.83 and you can do this by the add menu (Ctrl+A) and choosing Volume > Import OpenVDB. You can search around for VDB files to download or purchase.
Setting up the lighting
I’m trying to use the light and shadow shapes to draw us to the focal point and be as gratuitous as possible with artistic license. Whatever method and configuration of objects I can arrange to look good and to avoid confusion or distraction. Here I played with two concepts, one where we had some foreground light and one where the shaft of light was hitting the far building but pointing to the ship. I found lighting the foreground was too distracting and went with the second option. Here I also brought in some ground-plane elements as I was still thinking of doing it all in 3D. Later I’ll switch tactics and decide to make it all in Photoshop.
Adding main post-apocalyptic elements
I grabbed old cars from Sketchfab to populate the scene. I felt like it would be good to show cars submerged in water to give an indication of how deep the water is, but also show a little story-telling. We also have some broken ground pieces and our fire hydrant and of course the ground plane is now reflective to indicate the water. All the basic elements are there but at this stage I’m really concerned about how strong the visual statement is, and how much control I have (or don’t have) over all these ground elements including the grass. Let’s do a deep dive on grass next…
Grass in Blender
Okay, so it’s definitely worth taking a moment to point out that I did use Blender’s grass in the end. We’ll cover what I did for the final painting later, but I still feel like it’s useful to know how to do grass for shots like these, as there’s possibly some instances where it’s necessary.
- Step 1. Pick the object you want to scatter the grass on and go to the particles tab. Here you can see my base object and that I have scattered four kinds of grass onto it (in the top highlighted list).
- Step 2. Add a new particle system with the ‘+’ icon, choose ‘hair’ (rather than the default ‘emitter’) and click on ‘advanced’. Do this with as many species of grass model you want to scatter. Next you’ll choose the ‘render’ menu and pick ‘collection’ from the ‘render as’ drop down. You’ll see that I currently have large straight strands of hair sticking out of my base mesh. Once I pick a grass model (or ‘collection’ of models), this weird straight spaghetti will then turn into grass.
- Step 3. I won’t go into creating or sourcing grass blades right now, but you can find all kinds of free grass models, as well as fairly simple tutorials on how to make grass online. I had a few collections which contained different types of grass and even one with some little white flowers. Next I decided to use the ‘weight paint’ feature to paint on exactly where I did and did not want the grass. You can change the intensity so there’s thinner/smaller grass in places so it’s not all uniform.
Step 4. Here you can see the tab where you change your brush settings for the weight paint. You’ll also want to generally play with all the particle settings to get some variation, just like any other particle system or grass scatter plugin. To get the grass to only show where you’ve painted, make sure to look at the ‘vertex groups’ section of the particle settings and choose ‘group’ in the ‘density’ setting. You can also choose the same group for the ‘length’ setting and it’ll make the grass shorter at the same time as it’s getting less dense.
I watched an interview with one of the Last of Us artists, Florent Lebrun, where he talked about the ivy he used in some of his environment concepts. It blew me away to find out that he was using one of the most ancient ivy generators in the history of 3D ivy. It’s a standalone application literally called Ivy Generator and is from 2007 but still kind of works today. Emphasis on ‘kind of.’
Here’s a quick run-down: it’s free, it generates fairly realistic ivy, you import your building geometry, the geometry comes in all broken, as far as I know there’s no way to fix that, and so the ivy climbs over the broken geometry and you have to just deal with that. The photo below illustrates very typically what I’m referring to. Oh, and most of the controls make no sense and are back to front, so what you think turns something down, actually increases it. Man I wish someone would just update it a teeny bit. The other thing is that you often want the ivy to just keep going and spreading, but it will usually stop suddenly, so you have to create several chunks of ivy rather than just the one.
To use it, just export your geometry (I recommend making a basic approximation of your building, even if it’s just a cube with one or two extrusions that loosely follows the shape of the building) and then you simply import the geometry into Ivy Generator, double-click where you want the ivy to start and click ‘grow.’ Then you just hit the birth button to see the leaves. Simple. Well it’s simple until you want to control where the ivy goes so try to stick to defaults or you’ll soon end up with a tiny ball of ivy that won’t behave. Once you’re happy, just export it and bring it back into your 3D package of choice and apply the materials and opacity masks that come with Ivy Generator. In the image below you can see that I made my own approximation of the near building and grew the ivy on it. I did a couple more and brought them into Blender. I’ve also shown how it looks inside Blender.
Finishing up in 3D ready for Photoshop
Now we’ve made all that grass and spent ages tweaking it, let’s literally throw it away because it’s just not realistic enough. Hallmark of a mature artist, I’ve come to learn. I hear about it all the time when I talk to other artists – if you spend enough time on something, you’ll keep it in the picture, even if you know it’s not working. I’ve been there enough times to just bite the bullet. Let’s just go with photos. I knew I’d be heavily relying on reference photography (a dangerous path) but I have to be brave. So I deleted the entire foreground but also gave myself even more room to play with at the bottom of the render frame. I got to work rendering out my elements as per the previous tutorial and began assembling them in Photoshop.
I’ve got all my layers ready in Photoshop. Right now it’s not too many layers. I have everything from the ship forward on one layer and everything behind it on another and I’ve cut out the sky so I can just drop cloud layers behind those buildings with ease. I chose a sky that brought a degree of focus toward the general focal area of the building with the shaft of light and the ship. I then laid down some atmosphere with the soft round brush to make the ship stand out. You’ll also notice that I got rid of some of those spiky rubble shapes as they were too distracting.
Integrating the ivy
I found the ivy a little dark and also wanted it to spread out and interact with the building a little better. So I tweaked the levels (I rendered a mask for the ivy so I could isolate and adjust it) and then placed ivy photos over the top. I used a cool brush that was basically the shape of an ivy leaf which I then set to a slight scatter. I’d place down the ivy photo where I wanted it, put a mask on it, invert the mask to make it all disappear, then bring it back by painting on the mask using my ivy brush.
Here I also added some new photo-details over the building and made the ivy interact with those new windows. Let’s talk about those windows next…
Adding photo-realism to 3D models within Photoshop
Here’s some of the steps I used when adding photo-textures. This method has evolved over the years beyond just dragging in a photo and setting it to ‘overlay.’
Step one - Here’s the base photo. We have to skew it into place and color correct it and then we’ll use the ‘blend if’ feature to properly blend it into the layer below.
Step two - I’m now skewing it into the right perspective and I’ll also color correct it to be similar to the 3D render.
Step three - Next I’ll tend to set the layer to something like ‘soft light’ just so I can see behind the texture and this helps me drag the texture around to see where it’ll best interact with the base 3D render.
Step four - This is the rather cool ‘blend if’ feature within the layer styles palette. Here you can see an example of how I used to it add some slight wear to the concrete. There’s no hard and fast rules, just drag the sliders until you like what you see. Just remember to Alt+click on the arrows to separate them which creates a soft feather rather than the ugly hard result you’ll get if you don’t. Again, use masks to add and remove the areas you want to be seen/hidden.
Realistically integrate photos over your models.
This is the first steps towards adding photos to bring photo-realism to the ground elements as well as vegetation. I expanded the canvas at the bottom to once again allow for as much room as possible for the ground plane elements. I got some decent photos of trees from ArtStation’s Marketplace for only a few dollars and I set about cutting some of the trees out and placing them in.
At this stage, we’re essentially treating the piece like a matte painting and so everything needs to be as seamless as possible. I’m still not satisfied with the ground plane so I’m going to be spending a ton of time going through photos to see what I can find that’ll really draw us into and through the frame.
Oh, it’s probably worth noting that I experimented with a different sky here. Sometimes I’ll do that. I eventually return to the original though.
Leading the eye
So I spent ages looking through all the various sites to find a photo of a stream that would run diagonally and help lead the viewer’s eye from the right side of the foreground, through to the far middle ground. I found one at pxhere.com and it mainly served as a general placeholder for now. I’m also building up the vegetation elsewhere and slowly getting rid of much of the 3D from the lower half of the image. I’ve also used some thick low-lying fog to really reinforce the focal point and create a bit of mood too.
Nailing the reference
I kept searching around for better reference as the stream I found was just too noisy. Noise is a huge thing that you don’t really observe when you’re starting out in digital image making. When we look at an image, we really want to see something peaceful and harmonious where our eye is gently encouraged to move in a predictable manner through the space. This is achieved by deliberate and carefully placed points of contrast and/or interest. With noise (a noisy texture or photo for example), you’re accidentally distracting the viewer’s attention, creating dis-ease in their experience. I also gathered quite a few images of waterfalls before I found the exact ones to place in the distance. I wanted the water to have an origin. This was something I had in mind right from the start.
Here I’ve added lots of small things to bring the image to life. I’ve brought the fire hydrant back in, and also added moss to it, so it integrates with the mossy environment. I’ve added ivy onto the nearest car so it looks like it’s lived there for at least a couple years. I’ve put grassy lines to help direct the eye as well as a few little evidences of man-made building materials in the underlying moss. We have to remember the realism aspects – this isn’t a stream in a meadow, it’s a stream in a once-inhabited city. I also added a distant car because the scale of the ship looks questionable.
It’s definitely in-scale in terms of the scene itself, but it’s hard to appreciate how far back it is. The car helps to show that the ship goes back further than those two mid-ground cars. Some of the buildings have also had a little treatment to make them look a little more worn and overgrown. It’s subtle but needs to be done. If we saw more of their ground-level I’d be doing much more work along the lines of vegetation, but it’s all hidden in fog or trees.
At this stage you need to start considering all the elements of realism and believability and then making sure the added set pieces make sense.
The final grade
After adding a few more fine details like the birds and a little more suggestion of moss-covered rubble (vs moss covered pebbles/boulders) it’s time to step back and see how well the piece reads.
I wasn’t happy with the overall noise levels in the previous steps but I didn’t want to just turn the shot super dark. It was a bit of a back-and-forth trying to decide on details vs read. I ran the image past my friend and mentor Quentin Mabille who did a very quick paint-over and basically did what I was afraid to do – remove the details with a big soft round brush so we focus even more on the things we need to focus on.
If you compare this image with the previous step, you’ll notice the ground plane is a little too demanding of our attention. The image is generally a lot darker than I had planned from the start and now it’s like early evening with heavy rain clouds but one shaft of sun light. Originally I wanted quite a sunny day but it is very tricky to create a tight focal point under such lighting conditions. That said I’m happy with the result and hope you are too!
Credits - Sketchfab Models
- Sabri Ayes