Coherent Creature Design

As I promised in this chapter I will dig deeper into the designing process around the creatures that will be walking on the procedural terrain. Therefor I will present a technique, which is actually more of an artist’s life hack to keep yourself unintentionally creative than an actual technique in my opinion. To elaborate further on this, I made a tool which helps me to stay coherent in the way I design environments and creatures in this particular project. I’m referring also to two design books I’ve read through very rapidly.

So, grab a pen and paper – we’re be exploring the unexpected.

I bet those of you who are into drawing experienced this also once or twice while working on a piece. One is sketching a scene or character concept very rough and with a lot of gesture on a piece of paper (or digitally) but in the following steps, when it comes to outlining the shapes, the sketch becomes sort of stiff and loses its dynamic. The precise placement of the dark ink lines becomes the opposite of the strongly dynamic gesture you started working on. I realised pretty soon, when I did comics and cartoons, that I lose a lot of soulful characteristics and energy of the sketch when I “worked the illustration out”. Also, I realised some years ago when we did life drawing in class that I produced way more vivid results when we did ten second gesture drawing compared to a 50 minutes anatomy study. I guess it was not so much the time span, but the freedom of not being able to work the drawing out within my comfort zone. To capture the pose with a few fast and vibrant brush strokes or splatter watered ink without having full control of the end result, generated new and interesting shapes I could take as a base for later refining. In the following weeks, I started to investigate into gesture drawing in general to pimp my stiff handed illustrations.

The following video by Stan Prokopenko is a cool point to start with.

I still benefit from this change of mindset. Even when I do my favourite type of illustrations with a high, let’s say, “richness in detail” I start with very intuitive gestures for some unexpected spice.    

To increase the effect of not controlling one’s base sketch too much, I also began to draw with closed eyes. Sounds a bit weird, but I think everyone did this as a kid. I realised this for myself when trying to make a character for my game and the traditional way of setting up a skeleton (the sketches in the white box of the image) just gave me over and over again similar results. The other characters are made by starting with a blind drawn shape.

As you get better in drawing, I assume you will also get better in imagining things in front of your inner eye. In my opinion, and I am not a psychologist or neuro-scientist at all, the eye-hand coordination is kind of a brain-hand coordination that works also with closed eyelids. The tricky thing here is, if you train it well and the object or thing you imagine is not to complicated, you still get a decent and not totally unrecognizable result. But in this case, I rather get something totally out of any context so I can interpret forms and gestures into it. So, don’t think of a dog when drawing a dog with closed eyes – just scribble stuff and try to find a suiting shape.  Here are some more examples of these "fluid associations":

This associative approach is embedded in the human perception of the world. The effect that causes the “seeing faces in clouds” thing is called Pareidolia. It describes the phenomenon that people tend to see faces, or a better term would be “patterns”, in structures where are none.

I really like this approach because it opens a wide variety of interpretations on just one starting layer which can be elaborated more in detail later on. Further it’s not limited to visual appearances. As Wikipedia says also hidden messages in sound files that are played reverse or voices from random wind noise belong in the same category of slightly schizophrenic human interpretation.    

I found the following website, al.chemy.org, as a reference for concept artist which exactly belong to the topic above. I downloaded it just a few days ago to try it on my drawing tablet. It works properly, even if the pen strokes are translated a bit obviously sometimes – but this might be caused by not enough variables in my settings. I can imagine this as a great tool for artists of all kinds, even when it focuses on digital painting and concept art. It even tries to include dynamics by extruding the line waves. I suggest you try it out yourself. This image is made with the provided software.

I downloaded it just a few days ago to try it on my drawing tablet. It works properly, even if the pen strokes are translated a bit obviously and very flat most of the time – but this might be caused by not enough variables in my settings. I can imagine this as a great tool for artists of all kinds, even when it focuses on digital painting and concept art. It even tries to include dynamics by extruding the line waves. I suggest you try it out yourself.

Ideal type doesn’t mean necessary “the best possible phenotype”. It’s more a term describing the most common shape of something in people’s mind. This kind of bird shape represents the “default version” of any bird in western culture. Wings, not to small, not to big, medium long tail feathers, short neck, medium sized beak, thin legs with tiny claws. All the special features of other bird species are recognisable for us, because they differ from this ideal type. I guess that’s also why it’s easier to name a swan (or even just the silhouette) with all the extra shaping (long S-curved neck, huge white body, webbed feet) for children than to name a Phoenicurus, which is very close to the ideal of a bird.    

To create unique, or at least not very default looking bird shapes (in the context of art), blind shape drawing and associative structure recognition (these scientific terms tho) are in my opinion good ways. They increase the chance for not landing automatically in an ideal type of the desired creature. I also tried some morphing with a first sketch of a creature’s head I didn’t like much. It was way to unspectacular for what I expected it to be. I used the smudge tool in Photoshop and simply played around with it – way to controlled if I look back now. 

But what I got: Several new base shapes of a given structure (eyes, mouth, jaw) which I could turn into another creature within the same style easily. Morphing is also described in the book same book from Krippendorff as a way of elaborating the common shape associated with things. The example shows how layering all shapes of the different cups make out the ideal type for “cup”. I scanned the following image from the book. I put the credits on the end of the post.

Gunther Rehfeld, author of the german book “Game Design und Produktion”  from 2014, mentions the importance of shapes in his chapter about concept art. He writes about how players must be able to determine if a creature is a friend or enemy just by its shape and from long distance. The same is valid for weapons of all kind.

In order to use this great technique of blind drawing and its advantages over “just drawing monsters” I felt like I need a method behind all of this. One should be able to explain why a design appears the way it is. Like with the layering of versions to elaborate a default shape. Saying “I draw this church every day at the same time from my bathrooms window for one year to study lighting” is a well-made choice and method. But if you do it one day only with ink on paper but the next day with crayon on wood and then with a stick into sand it becomes a bit random and weakens the system.

What I want to clarify here is: I wanted a concept and an as much coherent methodology as possible behind it. Traceability is the buzzword. And I claim this is key in any “scientific” approach as well as on a master thesis. So, generating random shapes and fill them out from associations is not enough. My personal solution was to program my own shape generator, based on values from the terrain generator. This way, the creatures I interpret from the shape are based on the same ruleset as the terrain of the world they live in.

The application I made is a shape generator. It uses Unity’s LineRenderer and TrailRenderer component to generate randomized shapes within set value borders. This examples here are not valid ones for my particular project, but still worth a peek. I used to have randomized colours and a randomized line thickness. On the third image, I tried to programed a draw mode option to directly edit the shapes in the generation process.

In the end, the black and white shapes from gave me the most satisfying results.

The starting shape is a perlin noise that has a circular gradient applied on it. It’s basically the falloff generator from an early episode of Sebastian League’s procedural landscape generator. The goal is to generate shapes this way that are directly dependent on the e.g. the noise seed of the landscape generator. Also, the decisions on spawn heights for trees and the water level influence the shape in a certain manner. I also considered physics based placement of rocks from the last post as a form giving method. All the shapes are then saved as png-files inside a folder on the desktop. Here some results I like.

With this base, I further elaborated the shape by adding colours and details up to a finished creature concept. One first example is a crocodile-like monster that lives on the beaches of the alien world. It may wave its flag tail around to attract other individuals or to mark its territory. I’m open for inputs.

I want the creatures to fit in the world visually. The reason I don’t want to hop on the train of designing them like they could exist on this planet is that I don’t have enough capability to do all the checks for plausibility. In 3D modelling software, there is no limit to what you can do. A designer can create an elephant-sized turtle with legs thin as the ones of a flamingo. When I decide to include “biological correct evolution” into my project I would need to dig way deeper into science. Every decision on the design of the creatures then must be depending on these elaborated rulesets. Some games like the Swiss game Niche work with genetics to get to a coherent appearance. In my case I decide the body structure of each creature based on the generated shapes and colour scheme of the terrain.

The following images are the further elaborated shapes from the shape generator. The shapes as a coherent base for the artist’s own interpretation is working quite well so far for this project and it’s also a whole lot of fun. I am positive that the shapes bring up design solutions I wouldn’t have discovered when constructing the skeleton of the creatures “traditionally”. On the same time, while working on them, I again noticed how I fell back into designing animals I already know from memory. This could be a point for critique I am aware. But as creativity doesn’t mean “innovation only” but also reinterpretation and collage and so on this is okay. A bigger issue was somehow the fact that I always saw mouths, eyes and legs in the shapes. I even realised shortly after my decision on which shapes I want to work on that I might have chosen them unconsciously on these criteria. To not be bound to it I picked the third creature (the bug like one) randomly from the folder. Here are the examples I made.    

As I mentioned in my first post I use Cinema4d as my preferred software when it comes to 3D modelling. I admit having trouble with exporting textures and stuff sometimes in the past, but at the moment I think it does its job quite well. One thing I had to keep in mind was again the number of polygons I want to use for each model. This is also the reason why the models are not super smooth. I simply want to be able to work on proper conditions in Unity on my machine later on. The sculpted model was then exported as a fbx-file before I constructed the fish-like joint skeleton. The reason I exported it first was that I didn’t want to have the sculpt tag on the object as I knew from previous projects that it might cause some issues with UV maps later in Unity. An easier step would have been to simply transform the sculpted mesh into a new object in Cinema4d. The result keeps the same. I rigged the mesh according to the creature’s anatomy and tweaked around the settings for a while.    

The upper image is a rapidly made rendering in Cinema4D. I didn’t put much effort into it. I just wanted a first impression of a possible in game view.

Yesterday I presented the current state of the project to my class and I got really good feedback from two of my classmates.

A girl, she does mostly graphic design, asked me why I think everything must have a high traceability in design decisions. I realised for myself that there is a break in this thinking as soon as the designer or artist starts to work on the shapes. The base is as possible (same algorithm that creates the noise pattern of terrain and so on) but what happens to the shapes after - how the creatures will look in the end - is open for the individual design decisions of the person working on it. I now want to point that particular thing out, as I think, in times of No Man’s Sky and other games that use procedural generation to its finest, it is still the designer’s choice that makes the game stand out in the end. The human, professional decision on aesthetics and stylization makes the design appealing, not any of those seed numbers from code. The worst would be if a design in the end looks arbitrary. And that’s still a big issue with procedural generation for visual content.

When I was at Ludicious game festival in Zurich last January, I asked David O’Reilly, a great artist and animator who started working on games as well, how procedural generation influenced his work - I was mentioning Everything. As O’Reilly even wrote kind of a manifesto, it’s actually an essay, that you can download from his website and in which he proclaims “coherence” as the most valuable criteria. Which I agree on very strongly. His answer was that he yet doesn’t know how to cope with all the possibilities of procedural generation within a high grade of coherence, as the complex algorithm might always spit out arbitrary content in a way where the humans, at least in design, still must be the one instance of controlled filtering for valid, visual appealing results.

The other big critique came from an industrial designer who simply said he likes the blind drawn ones more than the procedural shapes. He highlighted the fact that I used the same side view of the creatures in the shapes I created by code. And he was damn right! I realised it and began to think about this. I guess the opacity of the brush strokes made the not-procedural shapes more three-dimensional. When comparing these two examples it becomes obvious how flat the procedural shape is, compared to the one I've drawn by hand:

I will start working on this problem with the next shape creature I create.

With this I close this topic of creature design for now and hope I could give you some references and insights :)

I'm looking forward to share the mechanic and story behind of the game in the upcoming post. I see you there and be sure to visit my Instagram profile where I announce new posts first.