The-Island-was-a-World_catalogue-digital

19 I also found an early GAN-based image generator called Artbreeder, and used all sorts of codes that ran on Google Colab. This was a technically cumbersome period. There were a lot of crashes and everything took a long time to calculate, sometimes a whole night (by the way, the "stable" in "Stable Diffusion" 3 reflects the pride in a software that doesn't crash). Luckily, the pace during the pandemic allowed for long periods of waiting and observation, and that's the pace of oil painting anyway, which I'm used to. These images, so similar to reality and yet so different from it, began to feed my paintings, and thus was born the exhibition "The Moon is a Mirror"* at Braverman Gallery (2022),** which was also the first poem I presented. The exhibition was arranged according to the poem, with each verse referring to one of the four walls of the gallery. The poem was partly composed of the titles of the paintings, which were hung and numbered in sequence, from 1 to 33. The sequence stemmed from the logic of the GAN's operation, which is different from the logic of generators like Midjourney, DALL·E, etc. There were no text-to-image generators yet, only image-to-image generators. The GAN's image chain has a kind of genetic logic in that each image can "give birth" to a new image from itself or by "mating" with another image or with several images. In GAN’s and other generators, each new image is the result of a statistical operation that extracts it from the latent space, where possibilities and hidden potential await the moment of calculation, the moment of the artist's wish. The creative space to navigate is vast, with many illusions and traps along the way." So how did it evolve from there? For a long time, I tried to give "The Moon is a Mirror" a form in video but the attempt failed for various reasons, mainly because I just didn't know how to edit video. Since that year (2022), increasingly surprising and user-friendly AI models have begun to emerge, led of course by the elusive wonder called "text-to-image generator", and even more astonishingly, "text-to-video generator" and "image- to-video generator." And so, while writingmy second poem, The Island was a World , I had at my disposal tools that allowed me to expand my means of expression. Alongside the paintings, I created images, animation clips and sound, which influenced each other and eventually created a world (for me at least). Over time, everything was woven into the video, using my basic Premiere editing skills, acquired the hard way… 3 Stable Diffusion is a deep learning, text-to-image model released in 2022 based on diffusion techniques . * https://www.oreneliav.com/mount-zero-202 0 ** https://www.oreneliav.com/how-to-disappear-completely