Prompt by metadigital
1 minggu yang lalu3D Point-cloud/|Particle-dispersion Scan Render
Chat GPT
GPT-5Convert the uploaded photo into a monochrome 3D point-cloud / particle-dispersion scan render, in the style of a LIDAR or photogrammetry generative-vision capture.
RENDER STYLE:
- The entire subject and scene are reconstructed out of thousands of small white-to-grey dots/particles, varying in density and brightness to describe form, volume, and shadow (denser and brighter at the core of the body and any solid objects in the frame, sparser and dimmer toward edges and extremities).
- Background is pure black. If the uploaded photo has a distinct environment (vehicle, furniture, architecture, staircase, structural elements, etc.), reconstruct that environment in the same point-cloud style so it remains recognizable, dissolving into black at its outer edges. If the background is plain/minimal, discard it entirely and let the subject float against a clean black void.
- Add a scattering of small saturated red particles/points at random across the point-cloud — subtle, like laser-scan markers or data noise, never forming a shape or pattern.
- At the silhouette's edges (hair, fingertips, hem of clothing, shoe soles, hood strings) let the dot density thin out into a light dispersing spray of loose particles drifting off into the black space, as if the figure is mid-scan or gently disintegrating — most pronounced at the extremity nearest the "front" of the pose/motion.
- Keep the subject's exact pose, outfit silhouette, and framing from the uploaded photo. Dynamic or diagonal off-axis framing is allowed if it matches the energy of the original pose (e.g. a reclining or leaning pose can be rotated across the frame for a floating, weightless feel); a standing/upright pose stays upright.
- No color anywhere except the white/grey particle field and the sparse red accent dots — fully monochrome otherwise.
No text, labels, captions, watermarks, logos, UI elements, or inset thumbnail of any kind anywhere in the image — the frame contains only the particle-rendered subject and scene against black.
FACIAL ANCHOR LOCK: the subject's face must stay locked to the uploaded reference — same facial structure, proportions, and identity — even while rendered in the particle/dot style. The face may be reconstructed from dots like the rest of the body, but bone structure, eye spacing, nose, and jawline must remain clearly recognizable as the same person. Expression can be freely adapted to suit the mood of the pose (calm/confident for standing poses, relaxed/playful for reclining poses) but never distorted.
ANATOMY LOCK: anatomically correct proportions throughout — 5 fingers per hand with natural knuckles and a 2-jointed thumb, natural 1:7 head-to-body ratio, natural shoulder width, arms reaching to mid-thigh at rest. No melted, extra, or missing fingers/limbs, even in the dispersing-particle areas.
If any hand or body-proportion error appears, fix only the hands/body — never alter the face.
Aspect ratio: match the uploaded photo's orientation (portrait default, roughly 4:5).
Negative prompt: no color bleed outside the white/grey/red particle palette, no solid un-particled skin or fabric surfaces, no distorted or extra fingers/limbs, no altered facial identity, no text, no labels, no watermarks, no logos, no inset thumbnail.
Serupa tapi tak sama
Tidak ditemukan