AI Art experience
- My first motivation for using AI is exploring a subject. With AI, I feel I can visually reduce concepts and represent complexity. I’m interested in the objects that man can produce to transpose something greater, something inaccessible. The process is similar to that of the fetishist who displaces his desire in a reduced object. The starting point for my pictures is always a human subject: social, religious, anthropological, environmental or psychological. Each image allows me to delve deeper into subjects that concern or challenge me, often with an introspective dimension; I see photography as a human science. - Next comes aesthetic research. AI is based on “archetypes”, common figures that the software can easily represent. In the end, AI gestures are fairly close to reality: we work with what we know, as if we only had real materials on hand. It’s more difficult to create unusual things, or you have to do it by hijacking archetypes. I'm totally fascinated by the creative freedom allowed by AI. But it's above all its limitations and the ways to get around them that make me start a new photo every time. For each visual, I create thousands of images until I manage to make the AI slip into something unusual but believable.
Personal experience
43 years old, living in Montpellier, former multimedia project manager, then advertising manager in a communications agency for over 10 years. I then worked as a photo retoucher for haute couture for 8 years and have been a product photographer and photo teacher since 2022. In 2023, I integrated AI into my professional activity, and immediately found an immense freedom that perfectly matched my creative aspirations. Since then, I have been working on a form of photography that I see as a documentary, a synthetic documentary with a slightly offbeat feel. Under the influence of my past as a product and advertising photographer, I always tend, even with AI, towards frontal photography and impactful images. My work reflects, I think, the concerns of a generation that has grown up in a model that no longer quite works, that needs to unlearn. A generation in tension between contradictory injunctions, between individual development and the general interest. A generation that has learned the importance of voting and integrated the need to possess in order to be happy, but which lives with democratic disillusionment and the limits of liberalism.
Unexpected thought
I have a feeling of the unexpected every time I start a new project with AI. I systematically go through a phase of doubt between the first and the thousandth prompt, before realizing that once again the AI is going to go off the rails and come up with a surprising reworking of my initial idea.
Description
Kintsugi Format 4:5 / 1186 iterations In Japan, the term “Kintsugi”, which could be translated as “gold joint”, refers to a technique for repairing ceramic objects. Traces of dish repair using vegetable lacquers can be found throughout Asia, but Japan has added the specificity of a gold finish. In Kintsugi, the aim is no longer to hide the traces of repair but, on the contrary, to enhance them with gold powder. The accident becomes the starting point for a new life, and the golden ramifications underline the object’s past. I wanted to see if this technique could also repair a divided territory. The Middle East was the most obvious terrain, with over a century of conflict, occupation and interference. The scars there seem to be inscribed forever, like gold in broken ceramics. This technique raises the question of the capacity of a territory to be repared, as well as the difficulty of recreating points of contact after decades of conflict.
Process
AI is based on “archetypes”, common figures that the software can easily represent. In the end, AI gestures are fairly close to reality: we work with what we know, as if we only had real materials on hand. It’s more difficult to create unusual things, or you have to do it by hijacking archetypes. After my theoretical research, I always start with a general idea of the mise en scène and a list of materials. I then propose an initial prompt to the AI, which is then refined over several days and thousands of iterations. My prompts are fairly precise, with around 200 words describing the different shots in the image, the set elements, the main subject, the time of shooting, the light, the colors and the focal length. I leave the AI as little leeway as possible on ideas or style; to this end, I neutralize all functions aimed at stylizing the image, and I make no reference to visual trends or artists. With the exception of secondary elements such as a background, I don’t formulate generalizations about the staging; instead, I give the AI a sum of precise elements and let it propose an initial arrangement, which I then refine by adjusting the prompt. This often results in rather unsightly, chaotic images. It’s as if, at the outset, we weren’t speaking the same language, but by dint of repetition, the AI integrates the words and manages to create a harmonious object. As you refine and refine the request, there usually comes a point where the prompt becomes too dense for the AI. Certain initial materials disrupt the image because the AI is unable to represent them correctly. I then purge the prompt of all parasitic requests. When the main elements of the image are more or less settled, usually after the first thousand or so iterations, I can focus on the mood and feelings conveyed by the image. Little by little, I fix the light, the hue of a sky, the attitude of a character and anything else that will convey the desired feeling. Statistics and chance also play a part in the process, with the most relevant prompts being repeated hundreds and hundreds of times to push the AI’s layout possibilities. This is truly a collaborative work. The last phase of the process, which is extremely satisfying but rather heartbreaking, consists of selecting a small number of really relevant images, and then, by comparison, keeping just one. Finally, the selected photo is enlarged on a second AI, its details fine-tuned, colorimetric processing and textures retouched for a large-format print.
Tools
Midjourney + magnific + Firefly + Photoshop
As a creative entrepreneur, Liri loves to envision, execute, and push new ideas and bring them to life. Liri is co-founder of OfffTlv, a global visual design festival made for the curious, and founder of the studio Salty Water.
This artwork earned my ‘golden ticket’ with its authentic appearance and the deep story it embodies.
Authenticity is a quality that sets it apart from typical AI-generated visuals. AI-generated art frequently evokes a surreal, imaginative style, blending elements in ways that, while artistically intriguing, sometimes stray into the territory of fantasy. This image, however, stands out by virtue of how closely it resembles reality with a smart twist that says it all.
The scenery immediately caught my eye and aroused my curiosity, as I felt that there was a subtle message here that I was relating to emotionally. A golden river is flowing, forging a path across dry soil that is full of grief. Even though the artist’s perspective might differ from mine, the piece resonated on a shared, human level. It wasn’t just about the hardship itself but about our collective hope for growth, for restoration. This artwork didn’t just tell a story; it connected me to my own, echoing an unspoken resilience I wish for.
