Scenery 3D: a 2022 photographic scene in the age of AI

In early 2022, I spent almost a week making one image. Two models, an imagined antique harem setting built in 3D, photographs taken from many angles, and a great deal of work to bring everything together. Looking back at Gwen 3D through my recent experiments with AI leaves me amused, amazed and full of questions. This photographic composite tells the story of a dizzying change in pace.

Scenery 3D: two models populate an entire scene

The idea was to photograph a duo and place them several times within a single setting. The many figures in the finished composition are the same two models photographed in different poses. We had to imagine their positions and relationships within a scene that did not exist in the studio.

The shoot was wonderful and deeply inspired. That is a happy memory of the project. The long technical adventure surrounded that encounter: building a space where those real presences could belong, with coherent colour and light.

Building a place for the photographs

At the time, I used image libraries such as Shutterstock as well as DAZ 3D assets. Characters and environments could be bought for a few euros. For this project, I started with a purchased 3D environment and worked on its lighting to produce a usable render.

That first result felt extraordinary to me. The columns, arches, marble and central platform offered a space in which to imagine the scene. I can see its limitations today. What I remember most is what it made possible at the time.

1 — The first DAZ environment render, prepared for the 2022 project. Work by Yohann — Yes Photographies.

Purple gels bring two kinds of light together

In the studio, I used purple gels to make compositing easier. Colour was part of the construction from the moment we took the photographs. I then had to work on the setting, its shadows and its tones so that it could accommodate them.

2 — Shadows and purple tones developed to accommodate the studio lighting. Yohann — Yes Photographies.

Next came the cutouts, made manually at the time, for each photographed pairing. I had to integrate the separate shots, position them and process the whole composition. The first assembled version shows this stage: the figures are present and a story is emerging, but work on the atmosphere is still underway.

3 — First assembly: the two models appear in several poses within the setting. Photographic composite: Yes Photographies.

Almost a week inside one image

Between buying the 3D asset, lighting the setting, making the cutouts and processing the image, I spent almost a week on it. I was satisfied with the result at the time. The final image carries that investment, even if a viewer has no reason to count the hours behind each figure.

I like showing these stages because they reveal what the finished image brings together. The setting and the people originate in separate spaces. Processing draws them into a shared atmosphere, with softer, more enveloping light and an almost painterly texture.

4 — The finished 2022 scene, after processing the whole composition. Photography and compositing: Yes Photographies.

What AI changes in my backgrounds today

I returned to this setting to see what GPT Image could do with it. To my eyes, the result is beautiful and astonishing. The first experiment retains the overall arrangement of the 3D background while introducing a different richness of light, reflections and texture.

5 — A recent GPT Image experiment based on the 3D setting: its structure remains, with reinterpreted lighting and textures. AI-generated image.

For the second, I allowed more freedom, without imposing the structure of the 3D setting. The place becomes another hammam: a pool, a different perspective, openings and reflections. These two images are recent AI-generated background experiments; they are not stages of the original 2022 composition.

6 — A recent GPT Image experiment without imposing the 3D background structure. AI-generated image.

In my current experiments, creating a background with suitable colours and harmonising the layers with it can take five minutes. Those are the operations I am comparing with my earlier workflow. The shoot, the idea, the decisions and the entire creation of a composite are not contained in those five minutes.

My small stronghold of reality

I find AI wonderful for creating backgrounds, and I have no wish to hide that enthusiasm. There is still an almost literary aspect to the work: writing what you are looking for, finding the words, defining an atmosphere. Above all, there is the idea. When making an image becomes this fast, it raises real questions about what comes next.

Amid that change, I hold on to a small stronghold of reality: the people. In Scenery 3D, they really were there in front of my camera. The poses, gestures and inspiration of the session happened. The place around them can be built, transformed and reinvented. That encounter remains what I care about.

Yohann — Yes Photographies