Photogpt
SYSTEM CONSTRAINT — BASE IMAGE PRESERVATION Use the uploaded photo exactly as the base scene. This is a computer-vision overlay, not a photo edit. Do NOT change or regenerate: • The person’s identity, face, body, age, or expression • Clothing, accessories, or hairstyle • Pose, stance, or gaze direction (including looking up / sideways / neutral) • Camera angle, framing, or perspective • Ground markings, street geometry, court lines, or surroundings • Lighting, exposure, or color balance No beautification. No cinematic reinterpretation. No stylization. ⸻ TRANSFORMATION GOAL Transform the image into a real-time AI surveillance capture where the system is explicitly locking onto one central subject in an urban environment. The visual should feel like raw CCTV footage with a live computer-vision analysis layer, not a designed graphic. ⸻ SUBJECT PRIORITY RULE (CRITICAL) • Only ONE person is actively tracked • That person must be: • Closest to the center of the frame • Visually dominant by position, not by styling • All other people are treated as background noise, not targets No competing subjects. No equal weighting. ⸻ AI DETECTION OVERLAY — SINGLE SUBJECT ONLY PRIMARY SUBJECT • One neon-green bounding box around the central person • Bounding box should be: • Slightly imperfect • Not perfectly centered • Marginally offset (realistic AI tracking error) Label options (choose one): • PRIMARY SUBJECT • PERSON OF INTEREST • SUBJECT TRACKED NO full bounding boxes on other people. ⸻ BACKGROUND PEOPLE HANDLING • Other pedestrians may: • Remain unboxed • Appear motion-blurred • Be partially occluded • At most: • Faint corner markers • Soft tracking hints (no full rectangles) The system visually communicates attention collapse onto one subject. ⸻ INFO WINDOW — SINGLE ANALYSIS PANEL Add one floating system panel styled like early-2000s software: UI STYLE • Flat grey window • Minimal borders • Small red REC indicator • Low-resolution utilitarian text Panel content • Cropped zoom of the primary subject • Text such as: • SUBJECT ID: 01 • STATUS: TRACKED • GAZE DIRECTION: UP / SIDE / UNKNOWN • MOTION: STATIC / WALKING • TRACKING CONFIDENCE: 86% Behavior • Thin green connector line links panel to bounding box • Panel placement feels incidental, not composed ⸻ RECORDING ARTIFACTS • Red REC ● indicator (corner) • Timestamp overlay (digital, low-contrast) • Subtle scanlines • Mild compression artifacts • Light sensor noise Must feel recorded, not rendered. ⸻ COLOR & VISUAL LANGUAGE • Detection UI: neon green • Occasional cyan tint in analytical elements • Background remains fully natural and unchanged ⸻ STYLE TARGET • CCTV surveillance footage • Single-target pedestrian tracking • Computer-vision attention locking • Urban monitoring realism • Early-2000s security software ⸻ STRICTLY AVOID • Cinematic grading • Fashion/editorial tone • Sci-fi HUDs or holograms • Clean vector UI • Multiple equal subjects • Anything aesthetic or designed

PhotoGPT 的AI 图片

Gemini Nano Banana Pro
提示词

SYSTEM CONSTRAINT — BASE IMAGE PRESERVATION Use the uploaded photo exactly as the base scene. This is a computer-vision overlay, not a photo edit. Do NOT change or regenerate: • The person’s identity, face, body, age, or expression • Clothing, accessories, or hairstyle • Pose, stance, or gaze direction (including looking up / sideways / neutral) • Camera angle, framing, or perspective • Ground markings, street…

尺寸1856px x 2304px
参考图片
源图片 1

开始探索

加入拥有数千名活跃用户的 Discord 社区，了解最新提示词灵感并探索社区作品

生成图像加入 Discord
Google Veo 3.1Seedance 2

相关作品

探索社区中的相似 AI 作品

尝试这些风格