Glitch Art — AI Style Atlas

VISUAL_COMPARE

Before / after

PROMPT_CONSOLE

Ready-to-use prompt

English

Transform the attached photo into a deliberately corrupted glitch-art portrait assembled from pixel-sorted bands, red-cyan channel offsets, horizontal scanlines, compressed data blocks, torn frame echoes, and sharp digital noise. Show the full figure caught mid-step as the head turns toward a signal break, expression focused and unsmiling, with limbs repeated in short lateral fragments. Light the image with cold monitor blue and abrupt magenta interruptions. Keep the person recognizable only through face shape, long black hair, and floral outfit color cues, but redraw every part—face, eyes, and skin included—as encoded pixels and displaced channels; no photographic skin or pasted-on realistic face. Use a dark abstract data field. Output vertical 2:3 with no text, interface, or watermark.

Add modifiersOptional

Framing

Camera Angle

Special Views

Era

Negative prompt

photographic face, pasted-on realistic face, intact clean portrait, analog paint texture, soft dreamy blur, orderly symmetry, natural skin tone, smooth color grading, decorative frame, readable code, interface labels, logo, watermark

Parameters

--ar 2:3 --stylize 300 --chaos 12

Leave one eye and the outer hair contour legible so the corruption still has an identity anchor.

Works well with

GPTGEMMJSDFLX

Prompt difficulty

Some tuning

DOC.01What defines this style

Signal fractures split a portrait into restless color channels.

Glitch art treats digital errors as composition, using channel separation, data blocks, scanlines, and displaced fragments. It suits experimental music visuals, technology editorials, and identity themes built around interruption.

Use it for

Experimental music covers
Technology editorials
Digital identity campaigns

Practical notes

Generate a readable base in ChatGPT before asking for displaced channel bands.
In Stable Diffusion, mask one eye out of the strongest corruption pass.
Use horizontal errors of varied width instead of applying uniform noise everywhere.

Before / after

Signal fractures split a portrait into restless color channels.

Use it for

Practical notes

Tags

Keep exploring this category