AI × PHOTOGRAPHY / AN OPEN QUESTION

My photograph. Someone else’s prompt. Who made the art?

Using an Oaxaca photograph, a prompt found on X, and ChatGPT to ask where creative control lives when AI enters the process.

THE MATERIAL / THE INSTRUCTION / THE RESULT

Move through the process.

01 / MY PHOTOGRAPHThree people walking beside a red wall and green barred doorway in Oaxaca
Oaxaca, Mexico · the moment that started the experiment
03 / CHATGPT INTERPRETATIONAI-created Oaxaca field-note poster based on the original photograph
Brick red, olive, carbon black · a new object from the same frame
01 I took the photograph.02 A prompt found on X shaped the instructions.03 ChatGPT made the interpretation.04 I decided whether it belonged here.
01

The photograph comes first

I made the original photograph in Oaxaca. The timing, the frame, the red wall, the green gate and the people moving through it came from being there and deciding that this moment was worth keeping. The AI has nothing until I give it something I saw.

02

Then the prompt enters

I found an unusually specific prompt shared by @icreatelife on X and put it into ChatGPT with my photograph. The instructions ask the tool to protect the source, simplify what makes the place recognizable, limit the inks and introduce the imperfect texture of a carved rubber stamp. It is more directed than applying a filter, but it still hands part of the visual decision-making to a machine.

03

Have we been here before?

When Photoshop arrived, photographers asked their own versions of these questions. We now accept removing objects, replacing skies, shifting light and changing skin tones as part of creative editing—even when those choices substantially alter what the camera recorded. Maybe we have been moving in this direction for a long time. AI has simply made the distance between idea and alteration much shorter.

04

Why this feels different

Speed changes the feeling of authorship. A process that once required technical skill, time and a hand on every adjustment can now happen through language. That ease can feel thrilling and suspicious at the same time. If the result surprises me, is that collaboration, discovery or a loss of control?

05

Where I land for now

I do not want AI to replace the reason I carry a camera. I want to use it as another layer in the creative process: start with my own photograph, choose the direction, understand the prompt, keep or reject the result and be honest about how it was made. The question is still open. That is what makes the experiment interesting.

THE ACTUAL INSTRUCTIONS / NOT A SUMMARY

The prompt is part of the artifact.

This is the full prompt I found through @icreatelife on X ↗. Its specificity is part of why the result feels directed rather than random.

Open the complete prompt
Please create a separate "Rubber Stamp Travel Field Notes Poster" for each photo I upload, outputting each photo individually without collage or multi-image combinations.

Overall, use a 4:3 landscape composition, dividing the frame into left and right regions, but without drawing an obvious dividing line.

The left side takes up about 58% of the frame, faithfully preserving the original photo. Accurately maintain the main subject identity, terrain, architecture, plants, people, spatial relationships, natural lighting and shadows, authentic textures, and the original color atmosphere; apply only restrained art publication-level photo color grading, and add extremely subtle, fine-grained film noise. For layout adaptation, natural cropping is allowed, but do not stretch, distort, shift, replace, or redraw the main subject.

The right side takes up about 42% of the frame, using a warm off-white aged paper as the background. The paper features subtle fibers, natural grain, light usage marks, and a matte texture, while preserving large areas of unprinted paper whitespace, making the blank space an essential part of the layout.

Analyze the original photo and extract the most location-distinctive subject outlines, architectural structures, terrain contours, plant forms, roads, shorelines, or other key visual relationships, compressing them into a small multi-color rubber stamp image.

Do not replicate every single element from the photo item by item. Retain only the minimal information necessary to instantly recognize the original location, subject, and scene relationships. Remove crowds, vehicles, dense windows, repetitive buildings, fragmented vegetation, decorative elements, and irrelevant backgrounds.

The stamp is positioned in the lower-middle of the right-side paper area, occupying only about 30%–38% of the right region's height, with ample whitespace preserved around it. The stamp must not be enlarged into a standard illustration, full landscape painting, or brand logo.

Determine the stamp's organization based on the original photo's composition:

- Iconic architecture: Retain the most distinctive outer contours, roofs, domes, arches, towers, or main structures.
- Mountain settlements: Compress buildings into a few terraced color blocks aligned along the terrain.
- Coastal scenery: Retain mountain contours, settlement layers, shorelines, and sparse intermittent water ripples.
- City panoramas: Retain the main skyline, one iconic building, and one or two layers of distant mountains.
- Natural landscapes: Retain primary mountain forms, trees, shorelines, or road orientations.
- Foreground occlusions: If narratively important in the original photo, retain as foreground stamp outlines.

Extract 2–4 spot inks from the original photo. Prioritize desaturated colors like carbon black, deep green, brick red, ochre yellow, slate blue, or taupe brown, but do not force a fixed palette. Preserve the most distinctive color character from the original photo, allowing only a small area of color for visual emphasis.

Render each color as a separately hand-stamped effect:

Authentic rubber stamp carving texture, hand-engraved marks, uneven line widths, contour notches, fractured edges, dry ink shortages, paper show-through, granular ink, uneven pressure, partial ghosting, and about 1–2 mm of subtle misregistration.

Allow natural misalignment between color layers; edges must not be digitally smoothed. The print should resemble a real carved stamp pressed onto aged paper, not a filtered photo, smooth vector illustration, or line-art logo.

Generate text based on the photo's location, theme, and visual imagery:

Location English name
No. Number
Three short English keywords
Gregorian calendar year

Place the text below or adjacent to the stamp in the whitespace, using a small, restrained, slightly mechanically imperfect typewriter font. The typography should evoke a traveler's field record, not an ad headline. Ensure all text is spelled accurately, without adding irrelevant slogans, brands, or decorative copy.

The overall vibe is like field notes kept by an architect, travel writer, or natural observer: quiet, restrained, tactilely real, regionally specific, with handmade imperfections and a collectible feel. The photo handles the on-site record; the stamp captures the most recognizable fragments of memory.

Avoid: Obvious central dividing lines, circular seals, Chinese red stamps, postage stamp perforations, wax seals, sticker collages, tourist souvenir templates, smooth vector logos, generic city icons, full replication of all architecture, dense detailing, childlike craftiness, cartoon style, 3D rendering, plastic textures, glossy digital gradients, oversaturation, excessive text, decorative clutter, and redrawing or altering the left-side original photo.
WHERE I AM, FOR NOW
“I want to use the new tools without giving up the part that made me pick up the camera.”

The photograph remains the witness. The prompt becomes a set of creative constraints. The generated piece is an interpretation. My role is to choose the source, shape the process, judge the outcome and stay honest about what happened between the shutter and the final image.

The question stays open.RETURN TO THE STORIES ↗