Skip to content

About

Turn documentary travel photography into restrained surreal cinematic tableaux.

Resources

Stars

290 stars

Watchers

0 watching

Forks

Repository files navigation

Reality Restaged

Documentary memory. Surreal stage.

A Codex skill for transforming documentary, travel, street, and human-centered photography into restrained surreal cinematic tableaux.

The goal is not to simply apply an AI style to a photograph.

Reality Restaged preserves the people, gestures, cultural details, relationships, and documentary memory of the original photograph — then rebuilds the world around them.

Real people. Real gestures. Real cultural memory. An impossible world.

⸻

What it does

Reality Restaged analyzes a source photograph before transforming it.

It looks for:

  • the visual hero
  • relationships between people, animals, and objects
  • culturally meaningful details
  • accidental documentary moments
  • the strongest visual contradiction hidden inside the image

It then develops a restrained surreal composition using principles such as:

  • cinematic staging
  • negative space
  • scale distortion
  • spatial folding
  • impossible architecture
  • controlled color fields
  • realistic human proportions
  • photographic film texture

The philosophy is simple:

Reality first. Surrealism second.

⸻

Visual Direction

Reality Restaged favors:

  • minimal cinematic compositions
  • strong negative space
  • theatrical staging
  • restrained surrealism
  • bold but controlled color relationships
  • realistic people and animals
  • subtle analog film grain
  • natural photographic lighting
  • culturally respectful reinterpretation

It avoids:

  • generic AI fantasy aesthetics
  • excessive surreal decoration
  • plastic skin
  • oversharpening
  • duplicated faces
  • distorted anatomy
  • unnecessary objects
  • repeating the same surreal device in every image

⸻

How to use

Give Codex a documentary, travel, street, or human-centered photograph and ask it to use the reality-restaged skill.

For example:

Use reality-restaged to reinterpret this travel photograph as a restrained surreal cinematic tableau.

Or simply:

Use reality-restaged on this image.

For concept exploration:

Use reality-restaged and give me three different surreal directions before generating anything.

For an existing composition:

Use reality-restaged. Treat this image as the master composition and only modify the element I specify.

⸻

Installation

Clone this repository into your Codex skills directory:

git clone https://github.com/traveler0621/reality-restaged.git ~/.codex/skills/reality-restaged

Then restart or reopen your Codex environment if necessary.

⸻

Core Principle

The skill follows one central rule:

One impossible thing.

Instead of filling an image with surreal objects, it looks for one strong intervention capable of changing the meaning of the photograph.

That intervention might involve:

Scale. Space. Shadow. Repetition. Gravity. Reflection. Material. Absence.

The result should feel less like an AI illustration and more like:

a real photograph of a world that could never exist.

⸻

Master Lock Mode

When an image has already reached a successful composition, Reality Restaged treats it as a master.

If the instruction is:

Only change the fish tail.

It should only change the fish tail.

If the instruction is:

Keep everything else exactly the same.

Composition, lighting, scale, color, subject placement, and approved visual elements should remain locked.

⸻

About

Reality Restaged began as a personal visual workflow for reinterpreting documentary photographs collected while traveling.

Rather than replacing documentary photography, the skill treats the original photograph as the memory from which a second, impossible reality can be constructed.

Documentary memory. Surreal stage.

Examples

Selected works created with Reality Restaged.

Reality Restaged — Selected Work

Reality Restaged – Selected Work

About

Turn documentary travel photography into restrained surreal cinematic tableaux.

Resources

Stars

290 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors