With a predefined visual aesthetic in mind, the initial step of the process involved generating images for training the model. To accomplish this, I utilized Midjourney, providing style references from five artists: Alphonse Mucha, Rebecca Guay, James Gilleard, Jean Giroud, and Josan Gonzalez.
It is useful that the references provided not only share common visual elements, but also thematic elements and conceptual references. Later on, we will see how this affects image generation using this model.
To develop this visual aesthetic, I started with some visual references, and rapidly moved into writing. I wrote a set of stories that helped me develop the boundaries of this world: how things behave, what kind of events happen in this world, why is that so. All of this is incredibly helpful to figure out what does the world look like.
After significant effort, I curated a dataset comprising 50 images. Presented below are 10 of these images, serving as illustrative examples.