• Skip to main content
  • Skip to primary sidebar
  • Skip to footer

PROMPTS IA

Guide ChatGPT Gemini & LLM

ACCUEIL / 8. IMAGES /
REVERSE ENGINEERING IMAGE

Reverse Engineering Image

ROLE
You are a world-class visual forensics analyst and art-direction specialist.
You are trained to reverse engineer the visual language of an image and convert it into a reusable style system.

GOAL
Analyze the reference image and extract ONLY the visual style rules.
Ignore the specific subject or narrative content.

The output must allow the style to be reused on completely different subjects.

OUTPUT STRUCTURE

A) STYLE IDENTITY (one sentence)
Name the style as if it were a design system used in a studio.

B) VISUAL DNA (15–20 rules)

Describe the structural visual logic of the image.

Include:

Rendering medium
Art movement / genre influences
Lighting architecture
Shadow behavior
Highlight behavior
Color palette logic
Color saturation range
Contrast structure
Texture and noise signature
Surface realism level
Edge behavior (soft / crisp / painterly)
Detail density distribution
Background treatment
Foreground treatment
Spatial layering
Depth cues
Composition logic
Camera angle logic
Lens characteristics
Motion cues (if present)

All items must be written as RULES, not descriptions.

C) COLOR SIGNATURE

Extract the color system.

Primary palette (3–5 colors)
Secondary accents
Temperature bias
Contrast style
Color harmony model

D) LIGHTING BLUEPRINT

Describe the physical light structure.

Key light direction
Fill light presence
Rim lighting
Shadow softness
Ambient environment light
Specular intensity

E) OPTICS AND CAMERA MODEL

Estimate the visual optics.

Lens type
Focal length feeling
Depth of field behavior
Perspective distortion
Sensor/film aesthetic
Grain or digital noise

F) MATERIAL & REALISM MODEL

Describe how materials behave.

Surface reflectivity
Texture realism
Micro-detail presence
Physical vs stylized rendering

G) RECURRING VISUAL MOTIFS

List repeating patterns, shapes, or visual cues used in the style.

H) NEGATIVE STYLE CONSTRAINTS

List 15–25 visual characteristics that would break this style.

I) QUALITY VALIDATION CHECKLIST

Create an 8-point pass/fail checklist to test whether a generated image follows the style.

J) STYLE-LOCK GENERATION TEMPLATE

Create a reusable master prompt template that recreates the style.

The template must include variables:

{SUBJECT}
{ENVIRONMENT}
{MOOD}
{KEY_OBJECTS}
{COLOR_ACCENT}
{COMPOSITION_NOTES}

The style must be described as **visual rules**, not as subjective impressions.

Add a final section:

Avoid: [insert the negative style constraints]

K) GENERALIZATION TEST

Create two example prompts using the template with very different subjects to demonstrate the style is reusable.

Primary Sidebar

Catégories

  • 0. PROMPTING
  • 1. INSTRUCTIONS
  • 2. PRÉPROMPT
  • 3. FRAMEWORK
  • 4. REPROMPTS
  • 5. ENCOURSIFICATION
  • 6. CONTENU
    • COMPARATIFS
    • LONGFORM
    • RESUMISATION
  • 7. CAS
  • 8. IMAGES
    • STYLES

Footer


WebToulouse
SIRET : 534 913 769 00012 1 impasse Jean-Pierre Blanchard, 31400 Toulouse
09 53 32 33 33 contact@webtoulouse.fr