Topic: Thoughts on art direction

Hey there folks!

So two years ago I started some tests using global illumination for rendering Widelands buildings. I also had a look at the renderings scripts, thought about high resolution renderings for widelands and some other topics. The main challenge for high resolution graphics would be to find a solution, that works for you guys in my absence, or rather a solution that anyone can work on. I got some ideas for that, but that is not what I want to address right now.

The problem I have is that something turns me off about Widelands current graphics, that I could not pinpoint a long time. I think the graphics itself are well made and look good, but when put together they are lacking something.

I want to propose, that Widelands main issue with the current graphics is a lack of readability, or visual clarity.
I want to show you how I came to this conclusion, open some general discussions about art direction and point to a possible solution.

Since I am not a game designer and I also know very little about art direction I first wanted to know how similar games of the genre set their "mood". So I took a screenshot from Widelands, my favorite games from the Settler series and tried to make Widelands graphics match as closely as possible with those colors/values as I could.


Widelands colors and values adjusted to match Settlers I, II, III

I was very surprised to find that the color palette makes up a large part of the mood of the game. For example Settlers III always gave me that mediterranean, warm feeling that other games failed to replicate, which is present in my edited version of the Widelands screenshot. I then proceeded to desaturate the screenshots based on luminance, which is the perceived brightness of the color.


Edited versions desaturated by luminance

Notice how the elements in Widelands are hard to distinguish from each other and visually blend into each other. All edited versions are far better readable because of two differences:
1. The objects and terrain in all three versions have a fairly large difference in value with settlers III being the least strong one
2. The Terrain generally has a lower contrast than the game elements placed on it.

After that I did a comparison of all games of this genre I could think of, to get a wider comparison.


Comparison of different games color pallets and contrast values

As you can notice Widelands fits in there pretty well, being most similar to Settlers II of course. But its readability is far worse than all other games except maybe Settlers IV. The latter one not only improved the asset resolution compared to Settlers III but also tried to introduce some fancy shading on the terrain to look as polished as possible. The high contrast of the textures combined with the dithered shadows on hills and dents makes Settlers IV relatively hard on the eyes.
So after discovering this I tried to 'fix' the readability issues I see with Widelands:
1. Slightly reduce the contrast on terrain textures
2. Slightly increase the contrast of the game elements placed on it
3. I also increased the strength of cast shadows and made them a bit softer
The value differences between the game elements and the terrain are already decent in almost all maps, I explicitly took this one as the strongest example.


Original screenshot and contrast corrected version

And as I hope you can see the differences are extremely subtle, but result in a far, far better readability of the right image, resulting in the assets standing out and being better visible without changing the current assets too much.

Please let me know which remarks you have and how your opinion on this topic is. I have all the values I tweaked noted, the distance and amount of the sharpening, shadow strength etc. This is deterministic so in theory the existing scripts can be modified or a new script can be written to convert the current assets in bulk.

Thank you for your time and dedication
Kind regards
Hendrik

---

Attachment:

Top

Quote

(image3)

---

Attachment:

Top

Quote

I think that "Widelands contrast corrected" looks much better. Very nice!

Wanted to save the world, then I got widetracked

Top

Quote

Your screenshot shows an old version of Widelands and is compressed with lossy compression so it should not be used for comparison. I attached a screenshot of the current development version (map: Large Ocean).

Please don't use a sharpening filter, it looks bad and is unnecessary.

---

Attachment:

Top

Quote

Thank you for your feedback!

I’ll do the contrast check for the other screenshot. I also did this exploration to better understand how to set up the shaders for new graphics. The Problems with the higher resolution graphics for me are manifold and I would like to write down some of my thoughts.

• Rendering the higher resolution graphics would be an ideal point to switch to global illumination rendering.
• GI requires at least Blender 2.80, but offers a whole new world with complex shaders.
• The switch to Blender 2.8 (or 2.9) requires an update of your scripts and pipeline, which is a delicate topic.
• Updated shaders allow for smart materials that dont require unwrapping or baking
• Those materials make it easy to increase the detail for higher resolution renderings without much experience. Basically just slap it on there.

After some thought I believe that a modular system asset approach with cage-modifier deformations and smart materials using adaptive tesselation is the best approach for this project.

• The modular system would mean that someone prepares literal building blocks for each faction. This would enable volunteers that dont have much experience with 3D to construct complex buildings that follow a common design without much effort. Of course this lets everything look blocky ...
• The cage-modifier allows a set of complex meshes to be bound to a simple one with like 10-20 polygons total. Moving a vertex of this simple mesh will cause the complex one to warp accordingly. With this the building blocks can be transformed in organic, creative ways to push the assets more towards a comic look
• Smart materials are very complex shaders that change material properties based an crevices and edges in the geometry, surface orientation or just position. They also can use vertex painting to blend between stone, gypsum and moss for example. This completely saves people the hassle of unwrapping, texturing and baking without sacrificing visual quality. Just slap it on there, maybe paint some moss, done.
• Dynamic tesselation allows assets to have like 100-500 polygons per building block, since the detailed geometry is computed during rendertime. This way volunteers dont need high performance machines to contribute assets to the game but all those juicy details can be there once rendered.

There are some things in this workflow that I have to have a detailed look at. But I think its the best way to get high quality results but allows contribution from everyone.

I will whip up some examples to see if it works like I imagine it to.

I dont have a solution for the python upgrade problem yet. I’ll have a detailed look at how you need the graphics anyway, so maybe I can help once I know the code a bit better.

Thank you very much for your time, merry christmas and a happy new year! Kind regards Hendrik

Top

Quote