Delirium Blooms is a work of software-based, generative art that creates unlimited variations on the same theme: sets of two flowers that are not representations of any real species, but pure algorithmic fantasies emerging from combinations of partially random shapes and colors.
This work does not use any kind of AI or machine learning technology. Rather, it has been hand-coded, line by line, fine tuning the right formulas and parameters to generate the different parts of the flowers, in a long iterative process of trial and error. It combines various resources from the software art toolbox, like particle systems, flow fields, boids, tweening, etc.
The starting point for this project was an invitation by Bright Moments Gallery to be part of their Buenos Aires Collection, presented at Palacio Guerrico from November 2 to 5, 2023. Motivated by that wonderful opportunity, I embarked on an investigation and development process that took several months.
Here is a description of that process, starting with things that inspired me and previous, related works; followed by a description of some technical aspects, the ever-important problem of color and an overview of different stages or progress.
Click on any image to enlarge it.
2. Inspiration
The main source of inspiration for this work is of course the endless wonder of shapes and colors in nature. But I have also been looking with intense attention at the way flowers have been represented in different moments and contexts in the history of art.
Plants and flowers have been a recurring topic in western painting for centuries, as many wonderful works and artists can attest. To pick just one example, I was recently fascinated by the discovery of the Dutch painter Rachel Ruysch (1664-1750), who was very successful in her lifetime for her exquisitely detailed representations of somewhat disheveled bouquets with insects crawling over them.
Eastern tradition’s approach to the subject is very different, but equally fascinating. For example, this flowers by Miao Guying (1875-1955).
In our times botanical art is a thriving scene with many interesting representatives. I’ve been looking with particular care and admiration at the watercolors of Fiona Strickland.
There are also great photography works that go beyond the simple celebration of the beauty of flowers to present very personal and touching visions. A case in point is Rachel Levy.
3. Related works
Delirium Blooms is a continuation of a series of generative works that I’ve been developing over the last couple of years. These figurative pieces take the artistic representation of nature as their subject and are experiments on the algorithmic simulation of traditional techniques. In this way, they are trying to build a bridge between software-generated images and the history of art.
A tribute to Anna Atkins (1799-1871), considered by many the first female photographer, Cyanotypes (August 2022) tries to capture some of the spirit of her pioneering work at the intersection of art and science.
The Likeness of a Flower (December 2022) looks towards expressionist painting, and assumes the challenge of capturing some of the looseness and fluidity of the human hand when holding a brush or a pencil.
Bouquet Numérique (March 2023) is a generative automaton specialized in making pencil drawings of flower bouquets. Strongly inspired by the astonishing work of British artist Claudia Lowry.
Before the Storm (May 2023) is my first incursion into landscapes, with a particular focus on the sky and the clouds, trying to obtain a resemblance to graphic techniques like wood engraving or etching.
4. Technique
Delirium Blooms is, essentially, a particle system driven by randomness. A large number of particles or autonomous agents drift around on the surface of the image, leaving behind colored traces that gradually build the drawing.
However, the behavior of the particles and the influence of randomness are highly constrained in several ways. We can describe the system by focusing on its three main features:
Hierarchical organization: Each particle can spawn children particles, and those can have children themselves, etc. That means both the initial positions and the subsequent trajectories of a given set of particles can be partially or totally determined by their parent particles. For example, the general shape of a petal is generated by a single, invisible particle, while the texture, color and detail come from a set of children particles that follow it.
Multiple forces: The movement of particles can be the result of the addition of multiple different forces. For example, one particle can be attracted to its parent, but also be influenced by a noise field and some random wiggle.
Tween system: Every property in the system can be adjusted through the life of the particle according to various formulas, like easing curves, sine waves or noise. In that way, it becomes possible to generate continuous variations in color, transparency, stroke girth or any of the factors that influence the forces operating on the particle.
A quick illustration of the above:
A single particle, moving left to right with constant speeed.
A single particle, moving left to right with constant speeed.
The same particle after a new force, driven by a noise function, has been added.
The same particle, now with children particles that follow it, at a distance controlled by a sine curve.
The children particles after a noise-driven force has been added.
Here the parent particles has been hidden, and the children are rendered with variable stroke weigth and color.
The particle system defines the structure of the drawing on the screen. However, it wouldn't do much by itself unless we visualize it in an interesting way. That's where rendering comes in. Each particle becomes a little brush with its own coloring and properties. The system has the following characteristics:
Many circles: As you can appreciate in the details below, everything in this work is made of very many little dots. I’m using just one p5.js drawing function: circle(). No lines or other shapes of any kind. This allows for fuzzy strokes and…
Variable size, color and opacity: All the brush properties can change in a very small scale (from one dot to the next), as determined by tweens or random variations.
Layers: Since everything here is strictly 2D, and I don’t have anything like a 3D model of a flower, the drawing is organized in layers, so things in the front are rendered over things in the back. Additionally, within each layer there is a time delay so, for example, petals at the back are drawn before petals at the front.
These are some close-ups from hi-res image exports (Click to enlarge).
5. Color
Color is an aspect of this work that demanded a lot of dedicated work to develop and fine-tune the combinations and balances until I felt they were right. The process started, color picker in hand, with the selection of colors from photos of real flowers.
I finally settled on 11 different palettes, which have an equal probability of being assigned to any given flower. Here are the swatches for each of them, plus an example of how they look when applied to a random flower.
The top line of 4 swatches represents the stops in the gradient for the petals. The middle line of 3 are the colors for the pistils and stamens, and the bottom line of 2 are the colors for their tips (more precisely, stigmas and anthers).
6. Process
This is a selection from the hundreds of screen captures I took along the development process.
6. Acknowledgements
First and foremost, a gigantic Thank you! goes to my partner, coach, and artistic advisor Gabriela Cárdenas, without whom this work would simply not exist.
And thank you of course to the amazing Bright Moments team, who made everything possible and were always ready to assist and make things easier for us.
