When and why to use this
When the team needs to produce new images that feel like part of the same campaign — without a photo shoot — we use ChatGPT with a fixed visual recipe. The master prompt defines lighting, optics, color, composition and emotional tone. The only thing that changes between generations is the scene, the characters and the action.
This guide complements — it does not replace — the system principles documented in Photography. The visual family and the line that links still live there.
Core principle
To achieve visual familiarity, always use the same reference image and keep the technical block of the prompt intact. Changes are concentrated in the narrative content of each new scene.
Must stay fixed: reference image · nighttime, premium aesthetic · 50 mm equivalent optics · shallow depth of field · cool color grading · observational composition · natural, subtle expressions.
Can change: activity or situation · number of characters · profiles, ages and diversity · professional roles · wardrobe within the palette · work objects · placement of the main subject.
Recommended workflow
A simple, repeatable process so the whole team generates images to the same standard.
- Start a dedicated conversation. Open a new chat or a thread reserved for this visual family. Avoid mixing it with very different graphic styles.
- Attach the reference image. Always upload the same base photograph before pasting the prompt. The reference delivers more consistency than a written description on its own.
- Paste the full master prompt. Do not remove the composition, camera, lighting, color, environment and constraint sections. They work as a single system.
- Replace only the variable scene. Describe who appears, what they are doing, who they are interacting with and what the context is. Be specific, but do not contradict the fixed style.
- Generate and assess the first version. Check the coherence of light, framing, focus falloff and color first. Then the accuracy of characters, hands, gazes and objects.
- Correct with targeted instructions. Ask for specific changes while keeping everything else intact. For example: “Keep the light, optics and composition; soften the smile and add more foreground blur.”
- Document the approved version. Save the final image, the exact scene text and the corrections applied. That is the record for future generations.
A note on camera data. Expressions like “50 mm” or “f/1.8” are visual instructions that steer perspective and depth of field. They do not guarantee real EXIF metadata, but they help keep the result photographically coherent.
How to write the variable scene
The quality of the new image depends on describing the situation clearly, without rewriting the entire photographic system.
Recommended structure:
Scene: [role of the main character] + [action] + [people they interact with]
+ [work object or context] + [emotional tone].
Ready-to-use examples:
- Product presentation — A Latina product director presents a mobile prototype to two designers as they review comments on a screen. Everyone looks focused, close and subtly warm.
- Research session — Two UX researchers analyze interview notes with a service strategist. One of them points to a finding on the monitor while the others listen.
- Creative review — A creative director talks with a designer and a copywriter at a worktable. They are comparing two printed visual proposals.
- Technical work — An engineering lead explains progress to two developers during a late review. Other people are working in the background, screens out of focus.
The right level of detail:
- State observable roles and actions: presenting, reviewing, talking, pointing, listening, comparing, taking notes.
- Define the relationship between characters: who leads, who listens, where the attention goes.
- Specify diversity, approximate age or appearance only when it matters for the campaign.
- Include one or two useful narrative objects: laptop, prototype, board, documents, monitor.
- Avoid crowding the scene with too many simultaneous actions or decorative elements.
A good scene is described in two or three sentences. The technical block already controls the aesthetic; the variable scene only controls the narrative.
Master prompt
Attach the reference image first. Then paste this block and replace only the text marked VARIABLE SCENE.
Use the attached image as the primary visual reference. Generate a new, fully
photorealistic image that preserves the same aesthetic family, especially in
lighting, depth of field, color temperature, observational composition,
nighttime office atmosphere and cinematic editorial photographic style.
I want a modern nighttime office scene that feels authentic, natural and human,
as if it were a photograph captured during a real moment of collaborative work.
The image must not look like generic stock, but like a premium, intimate,
sophisticated and observational photograph.
VARIABLE SCENE:
[DESCRIBE THE SITUATION, THE CHARACTERS, THE ACTION AND THE EMOTIONAL TONE HERE].
The characters should look relaxed, attentive, close and professional, with
natural and subtly warm expressions. Avoid stiff poses, visible stress or
exaggerated smiles.
COMPOSITION:
Horizontal 16:9 format. Medium or medium-wide framing. Immersive composition
from the point of view of someone sitting in the same office. The main subject
should sit near the visual center and be the sharpest point. Include out-of-focus
elements in the foreground, such as shoulders, heads or partially visible
screens, to create depth and a sense of real observation. Include other
colleagues or screens in the background with soft blur.
CAMERA AND OPTICS:
Simulate a photograph taken with a professional full-frame camera and a 50 mm
lens, equivalent aperture between f/1.8 and f/2.2, natural perspective, no wide
angle distortion. Precise focus on the face and eyes of the main subject.
Shallow depth of field, soft bokeh and organic sharpness with no digital
oversharpening.
LIGHTING:
Cinematic nighttime office lighting from linear ceiling fixtures and the ambient
light of screens. Generally cool or neutral-cool environment, with natural,
slightly warmer skin tones. Soft shadows, elegant contrast and a gradual
roll-off in the highlights. No frontal flash, no flat lighting.
COLOR:
Premium cinematic color grading in cool tones: deep blue, graphite, charcoal
gray, soft black and subtle violet or cyan casts. Natural skin, blacks with
detail and controlled saturation. Avoid garish colors and the exaggerated
orange-and-teal look.
ENVIRONMENT AND ART DIRECTION:
A modern, technological, active office, with dark windows, glass, lit monitors,
contemporary furniture and linear lights. Smart casual wardrobe in restrained
tones: black, dark blue, gray or muted purple. Avoid large logos and strong
prints.
FINAL RESULT:
The image should look like a premium editorial corporate photograph: realistic,
intimate, modern and sophisticated. Keep the same visual family in light,
optics, framing, focus falloff, color and atmosphere. The only thing that
should change is the scene described.
AVOID:
Stock look, stiff poses, exaggerated smiles, direct eye contact with the camera,
daytime setting, flat lighting, wide angle, plastic skin, excessive sharpness,
saturated colors, perfectly symmetrical composition, legible text on screens,
unreal anatomy, deformed hands or duplicated people.
How to iterate without losing the style
Corrections must be local and specific. Do not describe the whole image again on every attempt.
Correction formula:
Keep the lighting, optics, color grading, depth of field and overall
composition intact. Change only: [SPECIFIC CORRECTION].
Useful corrections:
- Expression — Soften the main character’s smile; they should look warm and calm, not elated.
- Depth — Add a more out-of-focus silhouette or monitor in the foreground to increase the immersive feel.
- Light — Darken the background slightly and keep the skin natural, warm and clearly separated from the cool environment.
- Composition — Move the main subject slightly toward the center and leave more visual air on the right.
- Naturalness — Have the characters look at each other rather than at the camera. The interaction should feel spontaneous.
- Wardrobe — Replace the light garment with a navy one without modifying the rest of the image.
What not to do:
- Change light, lens, framing, palette and scene all at once. The result stops belonging to the same family.
- Use vague instructions like “make it more professional” without saying what should change.
- Ask it to copy the exact faces from the reference when you only want to replicate the style.
- Accept an image just because the characters look right; it also has to pass the photographic coherence check.
Approval checklist
Before adding an image to the series, review these criteria in order:
- Overall aesthetic — Does the image feel premium, cinematic and observational?
- Light — Does a cool nighttime environment dominate, with slightly warmer skin?
- Optics — Does the perspective feel natural, with no wide angle distortion?
- Focus — Is there one sharp main subject with gradual falloff in foreground and background?
- Composition — Are there visual layers, overlaps and a sense of immersion?
- Expressions — Do the characters look relaxed, attentive and subtly warm?
- Color — Does the palette hold blue, graphite, black and cyan or violet casts?
- Environment — Does the office look active, modern and photographed at night?
- Realism — Do skin, hands, hair, clothing and objects look believable?
- Series coherence — Could this image sit alongside the previous ones without looking like a different campaign?
Approval rule. An image does not belong to the series just because it shows an office. It has to hold the camera language, the nighttime atmosphere, the palette, the depth and the human tone of the reference all at once.
File naming convention:
series_scene_variation_date_v01.png
Example: office_product-review_latam-team_2026-07-23_v01.png
Operational summary
The quick version to check before generating a new image:
- Attach — the same visual reference image.
- Paste — the full master prompt, without trimming the technical block.
- Change — only the scene, the characters, the action and the narrative objects.
- Review — light, optics, color, depth, naturalness and realism.
- Document — the exact prompt, the corrections and the approved final file.
Consistency does not come from repeating the same scene; it comes from repeating the same visual language.