The image below was pulled from AIKIZI's gallery of thousands of published decodes. It was selected not for its beauty but for its density. Every visual dimension is loaded. What follows is not a review or a critique. It is a forensic decomposition: the image broken into its constituent parts, every decision catalogued, every influence traced. If you have ever wondered what your Midjourney sref style reference code is actually encoding, or why one prompt produces a masterpiece and another produces noise, this is where those answers live.

Primary Evidence / Subject Photograph / 816 x 1456px

Repressed algorithmic hostility. That is not a tag. It is a psychological profile extracted from the convergence of pose, gaze direction, facial micro-texture, and lighting. The subject looks up through shadowed eye sockets. The chin is tucked. The gaze is directed upward. The lighting carves the face into geometries. Every one of those visual signals converges into a single reading: this entity is calculating something, and it is not benign.

The material inventory reads like a manufacturing spec sheet. The skin is not skin. It is matte stippled synthetic dermis, identified with high precision. The hair is not hair. It is fibrous bleached keratin spikes. Decode does not guess at surface-level categories. It identifies the specific material composition of every visible element, the way a crime lab identifies fibers under a microscope.

The micro-artifacts are where the forensics get granular. Halftone dot patterns and crosshatched shading are rendering techniques, not subjects. They exist at the level of the artist's hand (or the model's latent space). Circuitry seam lines and iris pigmentation speckles exist at the level of world-building detail. Decode catalogs both.

The palette classification alone is more useful than any Midjourney style reference code. Split-complementary tells you the exact color theory at work: a base color paired with the two colors adjacent to its complement on the color wheel, creating visual tension with a richer range than a simple complementary scheme. The swatch names are contextual: not "blue" but Steel Blue Skin. Not "green" but Emerald Ocular Glow. Every color is named for its function within this specific composition.

Four light sources. Triangulated. The key light comes from directly overhead, carving deep shadows into the eye sockets and under the jawline. The eyes themselves are a practical light source: internal ocular bioluminescence. The crimson backdrop is not a flat color but an atmospheric wash, itself functioning as a light source. And a rim highlight catches the crown of the head.

This is the kind of lighting analysis that a director of photography would spend minutes articulating on set. Decode produces it in seconds. For anyone using Midjourney, Stable Diffusion, or Flux to generate images, this is the difference between "cyberpunk portrait" and a prompt that actually specifies the four-source lighting setup that makes the image work.

Tension points are the fingerprints of a composition. They name the contradictions that make an image arresting. This subject has human freckles on synthetic skin. Living green light in eyes surrounded by dead shadow. Cool-toned flesh against hot red space. Every tension is a decision the creator made, consciously or not. Decode makes them all visible.

The cultural trace is where Decode operates as art history, automated. This image descends from three distinct traditions: Neo-Tokyo Industrial Design (the cybernetic anatomy, the manga rendering), Franco-Belgian Graphic Realism (the Ligne Claire linework, the Moebius influence), and East Asian Cyberpunk (the genre framework, the transhumanist themes). Three continents. Three decades. All converging in a single portrait.

The base layer is Showa Era Manga: Katsuhiro Otomo's Akira, the foundational rendering tradition. Layer 1 adds Ligne Claire, the Franco-Belgian clear-line technique perfected by Herge and Moebius. Layer 2 applies Industrial Cyberpunk Brutalism: the material world of exposed mechanical infrastructure. Layer 3 wraps it in High-Contrast Noir Lighting: the chiaroscuro that turns form into drama. Four layers. Four creative decisions. The complete genetic code of this image's visual identity.

This is what Midjourney's --sref codes compress into a single number. Style DNA decompresses it back into the full architecture.

Every exhibit above feeds into this: a prompt that captures not just what the image looks like but how it was constructed. The lighting setup. The material textures. The cultural lineage. The rendering technique. If you paste this into Midjourney, Stable Diffusion, Flux, or any generation tool, you get something in the same visual family. Not because you guessed the right words, but because Decode extracted the actual DNA.

One image. Seven exhibits. Subject profiled. Materials identified at manufacturing resolution. Color system classified by color theory. Lighting reverse-engineered from four sources. Emotional architecture mapped through tension analysis. Cultural DNA traced across three continents and three decades. Style DNA decompressed into four stacked layers. And a prompt reconstructed that captures the full genetic code.

This is what Decode actually does. Not tagging. Not captioning. Not a style number. A complete forensic decomposition of visual intelligence.

Every image in your library, every Midjourney generation, every photograph you have ever taken carries this same invisible architecture. The materials, the light sources, the cultural memory, the emotional tensions, the style layers. It is all there. You just need the right instrument to read it.

Submit Evidence →