#rachaelisnotreal

Starting May 2020, Rachael has been posting to the Instagram account @rachaellic one new image per day. She continued posting for 1000 times amassing a following of over 10K. Her capabilities are specially designed to create “likeable” images representing popular visual languages on the platform.

Rachael is not only a technical demonstration of the capabilities of real-time rendering, automation, and AI. Rachael intents to critically examine the consumption of contemporary imagery in the fast-paced social media landscape. Due to the continuously growing speed of how images are consumed on Instagram, the more successful (liked) images are the ones displaying an immediate, visually pleasing impression. On the contrary, images embodying deeper meanings or more complex ideas fail to be as popular, since they require more time to be seen and understood.

Announcement

Rachael has gained a decent following without any additional effort on my behalf. By posting similar, easily consumable creative assets on a daily basis she is building her following on a regular stable growth rate. 

What constitutes creative value? When I created Rachael my core interest revolved around reevaluating and analyzing cultural value in the digital art space. I intended to question if something is valuable because it received a lot of likes; if a large following defines a great artist; and if the algorithms of social media influence the way we express ourselves creatively as humans.

Explanation

Rachael generates images using various design methods and algorithmic techniques. The compositions are based on grid or dynamic systems and the placement of additional objects employs different techniques (randomization, trace methods to assess surface areas and general spawning probabilities). Due to its algorithmic nature, Rachael can generate an infinite number of images, each with a unique combination of shapes and colours. The image creation is primarily controlled by a weighted randomization of different parameters. The software contains a wide collection of objects that the algorithm can choose from, including different environments and lighting setups.

Rachael is a real-time software built with Unreal Engine 4 that generates images using various design methods and algorithmic techniques. The compositions are based on grid or dynamic systems and the placement of additional objects employs different techniques available in the engine (randomization, trace methods to assess surface areas and general spawning probabilities). Due to its algorithmic nature, Rachael can generate an infinite number of images, each with a unique combination of shapes and colours. The image creation is primarily controlled by a weighted randomization of different parameters. The software contains a wide collection of objects that the algorithm can choose from, including different environments and lighting setups. Rachael also uses Quixel Megascans assets for multiple props and shaders.

The software is fully RTX enabled and runs in real-time — the images are saved on the fly after the generator has finished building the composition. Besides Rachael’s core software even the process of uploading and posting is fully automated. The software generates a description for each image based on the selected shape and colour palette. A custom python-script python script (created by Julius Steinhauser @juliussteinhauser) posts it instantly to Instagram using a chance selection of tags. The post of a new image happens on a randomly selected time every day..

With the incredible advancements in machine learning and AI, visually pleasing images can be fully automated. If we neglect the communication aspect of visuals and simply use design techniques to attract likes, we as human designers risk becoming obsolete in the near future. Therefore, designers' priority should always remain finding smart conceptual ways of communicating our message, while the visual expression is utterly the vehicle that delivers it. Automating the human skill of visual communication that requires creativity and conceptualization is a much harder task to accomplish by an artificial algorithm.

In Action

The huge advancements of modern game engines have enabled these platforms to render near-photoreal CG imagery to a visual standard that comes close to the output of traditional pathtracers. All this happens in real-time with almost no latency between creation and outcome. Besides the immense improvements in visual quality and speed, game engines such as Unreal open up a whole new medium for 3D designers to explore. Rachael is a continuation of the exploration of alternative usages of game engines and strives to create a new approach to how we use and consume computer-generated imagery.

Credits

Design, Direction & Development
Matthias Winckelmann

Python Implementation
Julius Steinhauser

Sound
Chris Hoffmann