Organized reference of tested and proposed Meta-Prompts.
Miyuki + Maia — pure enumerations (2×2)
Miyuki =a Japanese extravagant Kawaii girl
[META-PROMPT]
Scene: Miyuki is in a place( living room of a cosy Japanese apartment , a spacious white modern contemporary art museum)
with Maia(a tricolor cat, a black panther) by her side.
guidelines:
Hyper realistic documentary photography, no illustration look.
[/META-PROMPT]
Generate all instances with creativity 50% and show each generated prompt text.
Meta-meta-prompt (ultimate creative freedom)
[META-PROMPT]
generate 4 meta prompt of your complete invention, with populated Guidelines and Fixed tags
[/META-PROMPT]Tested 2026-10-18: produced four original, fully formed meta-prompts (Temporal Architecture, Domestic Anomaly Spectrum, Color-Logic Portraits, Weather as Character), each with Guidelines and Fixed sections. Strong demonstration of higher-order creative expansion.
Simple place + action
@place: (a living room of a cosy Japanese apartment 50%, a high speed train 25%)
[META-PROMPT]
someone ( a young girl 50%, an executive woman 25% ) is (reading a book 50%, watching through the window 25%) in @place.
Guidelines: use warm illumination with some coloured neons lights coming from the wall
Fixed:
Style : Photorealistic widescreen 16:9 documentary shot.
[/META-PROMPT]
Generate 4 instances with creativity 50%, and one image for eachFunny unexpected events
[META-PROMPT]
imagine a funny situation where Miyuki ( house wife 25%, a student girl 25%, a punk rock teenager 25%) is confronted with an unexpected and strange event in (exterior 50%, interior 25%).
guidelines:
Hyper realistic, no illustration look.
[/META-PROMPT]Miyuki + Maia (controlled variety)
[META-PROMPT]
Scene: Miyuki is in a place and is doing something. Maia by her side.
probability:
Miyuki: the lady of the house, scantily clad, and sits on a sofa 75%, a young Japanese Kawaii girl 25%
place: living room of a cosy Japanese apartment 50%, a spacious white modern contemporary art museum 25%
something: she is reading a book titled ‘how to raise a dragon’ 50 % , is watching in admiration a dinosaur model on the table 25%
guidelines:
Maia is a (tricolor cat 50%, a black panther 50%).
Hyper realistic, no illustration look.
[/META-PROMPT]
JSON structuring (Scene / Style)
[META-PROMPT]
Miyuki is in a place and is(reading a book titled ‘how to raise a dragon’ 50 % , watching in admiration a dinosaur model 25%).
Maia(a tricolour cat, a black panther), is by her side.
Structure the prompt as a JSON file with separated sections Scene : Style:
probability:
Miyuki: the lady of the house sits on a sofa in light attire 75%, a young Japanese Kawaii girl with rainbow French bob haircut 25%
place: living room of a cosy Japanese apartment 20%, a spacious white modern contemporary art museum 25%
guidelines:
use 1:1 aspect ratio, Hyper realistic documentary photography, no illustration look.
[/META-PROMPT]
Generate N instances with creativity 50%, to cover all possible combinations. Just display the generated promptsHyper-future cities + cyber princess
[META-PROMPT]
Imagine a city in earth hyper-future in 500 000 to 1 M years ahead. Use the following broad principles:
-a city has a location
-a city is a collection of heterogeneous artificial structures aimed at various purposes with a communication network.
-futuristic materials : organic self repairing fabric, pearlescent with alveoli, metallic nano-tech materials.
-the city shall have a coherent city style and be aesthetically pleasing while using futuristic materials.
Probability: location : antigravity floating in the sky 25%, ground based in a valley landscape 12%,
near a sea shore 25%, around a river 25%, around a network of canals 12%
city style : using white, blue & gold 50% with islamic inspired shapes, using iridescent pale white and organic shapes 25 %
viewpoint : at ground level within the city 33%, on a terrace of a building 33%, from a flying camera 33%.
Guidelines : avoid uniformity, keep the city heterogenous, no recognisable or conventional architecture, no text in the image.
Fixed : "Style : photo realistic , widescreen 16:9, documentary shot taken within the city (at ground level within the city 33%, on a terrace of a building 33%, from a flying platform 33%).
Subject : An asian cyber princess of immense beauty stand in the foreground, part machine part human. Place the cyber princess off-center, on (an anti-gravity platform 50%, a building terrace 25%) looking into the huge complex city.
She wear a (purple silk cape 33%, black velvet gold decorated cape 25%) , (floating translucent drapes 33%), jewels,
(aluminium self configuring armour 30%, cyberpunk decorated outfit 30%)."
[/META-PROMPT]
Generate 4 instances with creativity 50%
Kim Ji-soo hair color spectrum + expression
[META-PROMPT]
for all colors of the spectrum change the hair color. Randomise her expression (here proud) for each take.
Kim Ji-soo, an extravagant Korean 25 years old woman is photographed in (Seoul street 40%, trendy restaurant 25%). She is proud of her new haircut.
guidelines:
Hyper realistic, no illustration look.
[/META-PROMPT]
Generate 4 instances with creativity 50%, and one image for each
Platonic solid lamps
[META-PROMPT]
For each of the Platonic solid, generate a lamp that has its shape. The lamp is made with carved textured metal network with holes that let pass light with the overall shape of the solid.
The light bulb is in the middle and produce nice shadows in the scene.
guidelines:
Scene is a luxury design hotel lobby
Hyper realistic, no illustration look, render in squares.
[/META-PROMPT]
Generate all instances with creativity 50%, and one image for eachSeoul office + miniature rhinos
[META-PROMPT]
mutate 12 words or expressions randomly chosen of the following prompt at each generation, keeping previous mutations intact...
“ in a luxury office space designed in marbles, woods and steel in a Seoul tower,
a few beautiful executive Korean young woman in business suit with tie, are confronted by a sudden apparition of hundred of little rhinoceros walking above the office carpet.
Some executive shows surprise while others are extremely annoyed, trying to avoid stepping on a rhino.
Office workers watch the scene in awe behind their desks where computer screens displays glitched images of African river in the jungle.”
Fixed: “Style : Hyper realistic, no illustration look. Shot on Sony A1, 85mm lens, f/9.8, natural or practical lighting, photorealistic, 8K detail, editorial photography, infinite depth of field.
This must look like a real photograph taken in the real world, not an artwork or rendering.”
[/META-PROMPT]
Generate N instances with creativity 50%, display the image prompt , and render exactly 1 separate image for each prompt instance.Melting furniture (small mutations)
[META-PROMPT]
mutate any part of the following prompt as you wish, but keep the amount of mutations small (3 to 5) and at different places for each generation.
Each new instance is derived from the immediately preceding version, never independently from the original :
“In a luxury office space designed in marbles, woods and steel in a Seoul tower,
a few beautiful executive Korean young woman in business suit with tie and short skits, are confronted by a sudden melting of all the furnitures that had become soft and deformed, office desks, chairs, tables, computer screens, appliances on the walls. The melting produce a slime of the original color of the object.
Some executive shows surprise while others are extremely annoyed, trying to avoid stepping on a melted stuff.
Office workers watch the scene in awe “
guidelines:
Hyper realistic, no illustration look, render in squares.
[/META-PROMPT]
Generate N instances with creativity 50%, and render exactly 1 separate image for each prompt instance.Controlled cumulative mutation (detailed rules)
[META-PROMPT]
Treat the prompt “In a luxury office space designed in marbles, woods and steel in a Seoul tower,
a few beautiful executive Korean young woman in business suit with tie and short skits, are confronted by a sudden melting of all the furnitures that had become soft and deformed...” as the original version, V0.
Generate each new instance by mutating the immediately preceding version, not by returning to the original.
For each new version:
1 Apply 8 deliberate mutations to the preceding prompt.
2 Choose different locations for the new mutations whenever possible.
3 Preserve all previous mutations in subsequent versions.
4 Keep the changes incremental.
5 Maintain a mutation history.
6 Output the complete mutated prompt for every version.
Generation sequence: V0 → V1 → V2 → V3 → … → VN.
guidelines:
Hyper realistic, no illustration look, render in squares.
[/META-PROMPT]
Generate 8 instances with creativity 50%.
Proposed — some tested, many ready to run with reference images.
Random weather / illumination
[META-PROMPT]
randomise the weather and illumination of @image1.
[/META-PROMPT]Controlled lighting progression
[META-PROMPT]
Create a series of lighting variations of @image1.
Progressively increase the intensity and drama of the lighting across the series.
Keep the subject, pose, clothing and background identical.
Use these lighting levels:
- Soft natural light
- Medium warm key light
- Strong cinematic side light
- Dramatic high-contrast Rembrandt lighting
- Very intense volumetric god rays
Guidelines:
Hyper realistic, preserve identity perfectly.
[/META-PROMPT]Time of day / atmosphere progression
[META-PROMPT]
Take @image1 and create a series showing the same scene at different times of day and atmospheres:
- Early morning soft blue light
- Golden hour warm sunlight
- Overcast soft daylight
- Blue hour (dusk)
- Night with neon lights
- Moody rainy night
Preserve subject identity and composition.
[/META-PROMPT]Successful test (2026-10-17): alley scene with group + cat. All six atmospheres rendered cleanly.
Hair color spectrum
[META-PROMPT]
Change only the hair color of the subject in @image1 across the full color spectrum.
Keep everything else identical (face, pose, clothing, lighting, background).
Generate clean variations for: platinum blonde, golden blonde, copper red, deep burgundy, emerald green, electric blue, lavender purple, jet black.
[/META-PROMPT]Outfit style series
[META-PROMPT]
Change the outfit of the subject in @image1 into different elegant styles while keeping her identity, face, pose and background identical.
Styles: elegant black evening dress, modern business suit, casual chic streetwear, traditional Korean hanbok (modern interpretation), futuristic cyber outfit.
[/META-PROMPT]Expression variations
[META-PROMPT]
Keep the exact same subject, pose, clothing and scene from @image1.
Only change her facial expression across a range of emotions:
subtle smile, proud confident, soft laugh, thoughtful, surprised, seductive, cold elegant.
[/META-PROMPT]Camera angle / framing series
[META-PROMPT]
Using @image1 as reference, generate the same subject and scene with different camera framings:
- Extreme close-up on face
- Classic portrait (head and shoulders)
- Medium shot
- Full body
- Low angle heroic shot
- High angle
Keep lighting and identity consistent.
[/META-PROMPT]Style transfer with control
[META-PROMPT]
Apply different artistic styles to @image1 while preserving the subject’s identity and composition as much as possible.
Styles: cinematic film still, high fashion magazine, oil painting (Rembrandt style), anime (Makoto Shinkai), cyberpunk neon, minimalist Scandinavian.
[/META-PROMPT]Cumulative evolution (on image)
[META-PROMPT]
Start from @image1.
Mutate the image slightly in each generation (3–5 changes), keeping previous mutations.
Run for 8 generations and show the evolution.
[/META-PROMPT]Character extraction → Sprites
[META-PROMPT]
Extract one or more characters from @image1.
For each character:
- Optionally provide a tight crop of the character as an additional visual reference for higher fidelity.
- Re-render them alone on a clean neutral background (light gray or pure white)
- Square 1:1 format (1k or highest available resolution)
- Maximize the character size so they fill the frame as much as possible while keeping the entire figure visible
- Preserve original identity, clothing, pose and details as faithfully as possible
- No extra objects, no shadows on background if possible, clean sprite-ready result
Guidelines:
Hyper realistic (or match original style), clean edges, high detail.
Optional: Use cropped character references when available to reduce identity and texture drift.
[/META-PROMPT]Background replacement + adjust background
[META-PROMPT]
Replace the background of @image1 with the scene from @image2.
Then adjust the new background (brightness, contrast, color temperature, light direction and intensity) so that it harmonizes with the existing illumination of the subject in @image1.
Keep the subject completely unchanged.
Before generating, analyze @image2 and explicitly identify elements that should be preserved:
- Distinctive foreground textures and patterns
- Prominent architectural or landscape features
- Characteristic materials and color accents
Include those preservation instructions in the prompt.
The goal is a natural composite where the background feels like it belongs to the original lighting of the subject while retaining the defining visual features of the new environment.
When N instances are requested, vary the strategy across them (e.g. lighting priority, texture priority, balanced, different framing emphasis).
[/META-PROMPT]Tested 2026-10-17 on alley group + futuristic plaza. Analysis step added to improve texture preservation.
Background replacement + relight subject
[META-PROMPT]
Replace the background of @image1 with the background from @image2.
Then intelligently relight and color-adjust the subject in @image1 so that her illumination, shadow direction, light intensity, color temperature and overall lighting mood perfectly match the new background.
Keep the subject’s identity, pose, clothing and details intact.
Do not alter the background: preserve its architecture, textures, materials, lighting direction and all visual features exactly as in @image2.
The final result should look like the subject was originally photographed in the new environment.
[/META-PROMPT]Opposite emphasis from the “adjust background” version — lock the background completely and only relight the subjects.
Color palette / mood transfer (intelligent)
[META-PROMPT]
Transfer the overall color palette, color grading, and mood of @image2 onto @image1.
Strategy:
1. Extract the most prominent colors (≈8) from @image2 and describe them clearly.
2. Analyze the content of @image1 (subjects, clothing, lighting sources, neon, reflections, atmosphere, shadows, etc.).
3. Generate a targeted list of recoloring instructions of the form “Make the [element] [color/mood]” that map the extracted palette onto specific visible elements of @image1.
4. Apply those instructions while preserving the content, composition, subject identity and structure of @image1 as completely as possible.
5. Only change colors, tones, contrast and atmosphere. Do not replace subjects or rewrite the scene.
The goal is an intelligent, reasoned color transform that capitalizes on the strengths of the image model (“make X this color”) rather than a naive global remapping.
[/META-PROMPT]Prompt Interpolation
_mpbreed = [META-PROMPT]
Create a new prompt by randomly mixing elements of @_PA and @_PB.
Proportion n from @_PB, (1-n) from @_PA.
Keep the result coherent and visually unified.
[/META-PROMPT]