ParisPictureClub
20 passage Saint-Sébastien 75011 Paris
contact@parispictureclub.com
@parispictureclub
contact@parispictureclub.com
@parispictureclub
2023 to 2026 - Dust
2023 - Reproduction through our standard 3D software
Reproducing a photograph as an exercise in decomposition and analysis. Chosen for its aesthetic appeal, its subject matter and admiration for the photographer’s work.
The process allows us to reverse the construction of a conventional image, where framing, composition and style are infinite and generally stem from an imaginative and creative vision. Here, the constraint remains reproduction, and every placement or adjustment serves to refine the resemblance to the existing source material.
At the start of 2026, with the advent of new tools based on artificial intelligence, trials were carried out to make specific improvements. At that point, we used Gemini solely to explore its capabilities, without considering it to be the best tool. These minor adjustments help to reduce the rigidity of certain elements. The original base, however, remains largely unchanged. Given the number of AI generations required, as well as the very small sections retained and assembled using masks during post-production on the image, you need to be patient. The full generated image never matches what you envisioned. I suppose you can be satisfied with the result if you have nothing precise in mind nor too many expectations.
AI describes the photographer’s image (with Florence-2-Flux) :
"The image is a frontal photo that shows a large, rectangular building with multiple levels. The building appears to be made of concrete, with a light grey color. The top level has multiple windows, each with a dark green tint. The windows are arranged in a grid-like pattern. Below the windows, there is a concrete wall with a diamond pattern. In front of the building, there are bare trees with no leaves. The ground is covered in a layer of snow. The sky is overcast, with some clouds visible.”
The overall mood of the image is bleak and desolate." AI generates an image based on its own description (using Flux-Schnell and Flux-Klein) :
"The image is a frontal photo that shows a large, rectangular and abandoned building with multiple levels. The ground floor of the building consists of concrete piles and a few concrete modules, the first floor has strip windows, the second floor has five square windows divided into two sections, and the top floor is set back from the main façade and has strip windows. All the glazing is missing, the window frames are made of wood and, in some places, turquoise tarpaulins have been fitted in place of the panes. The building is made of precast concrete panels with a visible grid pattern, with a light grey color. In front of the building, there is a boundary concrete wall with a diamond pattern and also bare trees with no leaves. The ground is covered in a layer of snow with a few dead blades of grass. In the background, you can see bare trees and a light mist. The sky is overcast, with some clouds visible. The overall mood of the image is bleak and desolate."
Result : not exact, but fairly close. The prompt could be refined as the model generates more outputs.