Week 6 - Whats that over there?
This has been one of those 'little bit of this, little bit of that' kind of weeks. I've made some more refined strides towards completing this scene even though it is still a long way off. As it gets a little more in depth this week, I thought I'd start off with something a little more simple.
"I made a rope texture!"
This is the first texture in this project that I completed entirely inside of substance designer which is awesome because it gave me some time to further my knowledge of the programs whilst picking up some tricks that I could use in future textures. I generally really love the hands on approach of creating tiling textures within Zbrush however for efficiency sake in the future, I would like to get to a comfortable place with designer that I could create textures in just the one program. This isn't something I am going to have to worry about in this project as I am currently doing fine with the Zbrush to designer workflow that I am using but it's still a nice area for gradual improvement for me. As far as this texture, I feel its simple enough that it reads well at a lower resolution as it doesn't have to be huge and it sells itself as a rope texture. I may make some variations or edit the material in engine to vary the texture but I will see how it looks in engine before I start on that.
I've also been messing around with a simple colour randomisation blueprint as well this week to test on my water Lilly pads. This set means that I can choose a set of gradients that I want the object to conform to and it will randomly choose between them based on world position. All I have to do then is duplicate and place them along the path to the tree and they will randomly change colours when I move, scale and rotate. I may apply the same techniques to a couple of other more basic areas of the scene to break up the uniformity of everything later on.
Keeping with the theme of textures and materials. I decided that the tree needed way more variation in its textures and layers of detail. To achieve this effect I first created a new tiling texture of bark to go on the roots of the tree with the intention of letting areas of bark come through when applying the moss back on later. I wanted this texture to be darker because of the added moisture from the water as well as for the purposes of providing a more natural gradient going up the tree to act as a com-positional element of the scene. I then started vertex painting my new and improved moss texture. After some experimenting with more detailed moss, I decided to keep it reasonably simple as it breaks up the visual complexity of the tree better as well as reading more like real moss from a distance. I am still going to be tweaking this texture as it stands but its definitely getting better with each iteration so only time will tell on this one. To break up the smooth silhouette that the moss was getting, I opted for some adaptive tessellation within the material using the height map I created in substance designer to get this more interesting silhouette that simplifies for performance sake the further you get away from it. This is an effect form of LOD-ding a texture because it doesn't need the visual complexity from a distance as it would up close so its just wasting performance costs. You can see this effect more clearly on the second screenshot. To take the vertex painting further, i can now use the material graph I created on any asset I want to give that mossy touch to as I'll talk about more next.
This is one of the assets that I added some moss too as it would naturally collect around an object that has been integrated with environment for a long enough duration. I slightly changed the material for variation sake also from the tree by changing the normal mapping settings. Instead of collecting on the surface so much that it takes over the shape after a long time, I decided to keep a softened version of the normals from the boat in the moss so that you can still see all of the gritty wooden boat texture below. This effect allows me to get more use out of the assets that I am creating without worrying about the viewer noticing obvious repetition of models because each model can be edited appropriately to where they are in the scene.
To change the subject matter a little, the lack of depth in the tree line has been a bothersome feature for me this week as its meaning that the scene isn't selling the depth that I am after. As you can see below, when you walk to the edge of the tree at the back, there is currently a never ending world of nothingness being shown. To break up the horizontal line from the floor plane, I dropped in some 3D land masses which solved the one issue. However, I needed to find a more economical way of implying a tree line without placing more of my quite performance heavy trees that are currently in the level. The way that I went about this effect was to create a 2D tree line planes that would be set to a lower opacity as it gets further back. This would mean that I can place as many as I feel I need without worrying about them costing a lot because they are far cheaper than the trees I currently have.
The effect of this can be seen below. I think its definitely doing what I am wanting it to do but I am now noticing that there is too clear a differentiation between the 2D and the 3D trees in the areas that the two meet. I've started looking into ways of overcoming this problem whilst keeping in mind that I want to lower performance costs of each layer. I'll talk about this a little lower.
The first technique I'm using is creating Lower resolution models for a more distance layer of trees and removing the opacity channel from them so they aren't adding to my alpha overdraw which can never hurt. Also, from the distance the player is at, they can only make out silhouettes at this stage anyways which is great.
I'm not sure that I'll keep this yet but I have been looking into flat shading the further layer of trees so that they visually start to blend with the trees behind them. The reason that I probably wont keep this however is that I noticed that because this tree is no longer being shaded to add to its three dimensional qualities, it was reading as 2-Dimensional. This meant that it may be an expensive triangle cost that turned out to be completely unnecessary and could be accomplished with 2D planes, so I'll look into implementing that more next week to see if its more or less effective than what you see below.
In other news... after some solid feedback from my tutors and peers, the scenes lighting has been gradually getting darker and darker meaning its far harder see any atmospheric perspective or general depth in the scene. As you can see in the images below, the trees in the background, the main asset and the foreground element all use the same level of values which is the exact opposite of what I am after. To overcome this in the coming week, I am going to do a decent lighting pass on the scene, bring up the atmospheric fog and control my point lights much more and get a solid lighting for everything that is currently in the scene. This should mean that all my values read better and there will also be no areas where complete blacks are visible because its currently looking too harsh and too unnatural. With this in mind, I'm heading back to work!!
