Video Prompt Reverse Engineer

Auto Video Prompt Reverse Engineer v3.0

Advanced AI video prompt reverse engineering. 输入 video/screenshot/description → structured prompts + reproduction 工作流.

When This 技能 Triggers

User provides any of:

Video link (Bilibili, YouTube, TikTok, Xinpianchang, etc.) Video file or screenshot(s) Text description of a video's visual style 请求 to analyze, deconstruct, replicate, or reverse-engineer a video Analysis Rules — Every Shot Must Include Dimension Required Analysis Shot Type ECU / CU / MCU / MS / MWS / WS / EWS Camera Movement Static / Pan / Tilt / Dolly / 追踪ing / Crane / Handheld / Orbit / Zoom / Speed Ramp / Snap Zoom / Whip Pan Composition Rule of thirds / Centered / Symmetrical / Leading lines / Dutch angle / Low angle / Over-shoulder Lighting Natural / Studio / Neon / Volumetric / Rim / Backlit / High-key / Low-key / Spotlight / Strobe / Muzzle flash Color Palette name, color temperature (warm/cool), contrast curve, saturation level Subject Motion Walking / 运行ning / Dancing / Falling / Turning / Slow motion / Freeze frame Depth of Field Shallow / Deep / Rack focus Texture Film grAIn / CGI / Ultra rea列出ic / PAInted / Pixel art Temporal Normal speed / Slow motion / Time-lapse / Freeze frame Style Identification 检查列出

Must identify ALL that 应用ly:

Director style: Nolan, Villeneuve, Refn, Deakins, Wes Anderson, Snyder, etc. Film genre: Cyberpunk, Atomic Punk, Film Noir, Neo-western, Post-apocalyptic, etc. Animation style: Anime, cel-shaded, 停止-motion, etc. CG style: Photoreal, stylized, low-poly, etc. Commercial style: Product hero, lifestyle, fashion, tech reveal AI artifacts: Temporal flicker, morphing faces, smooth physics, perfect lighting, etc. Color grading: Teal-orange, bleach bypass, film noir, vaporwave, golden hour, kodachrome, etc. Lens style: Anamorphic flare, tilt-shift, bokeh characteristics, focal length Film stock: Kodak Portra, Fuji Pro, Kodachrome, etc. 模型 Estimation

Identify likely AI 模型(s) used:

Kling: Smooth motion, Chinese prompt friendly, good physics 种子ance (小云雀): Audio-visual 同步, immersive short film mode, character consistency 运行way Gen-3: Camera controls (Pan, Zoom, Roll), cinematic 质量 Veo: Advanced cinematographic natural language understanding Sora: Narrative prompts, long duration, complex physics Pika: Short 命令行工具ps, motion parameter control SVD: Image-to-video, motion_bucket_id control Wan/CogVideoX: Chinese-优化d, shorter 命令行工具ps Midjourney/Flux: Keyframe generation (not video) Prompt Reverse Engineering Template

For EACH shot, 输出:

Shot XX

Content: [what's h应用ening] Camera Language: [shot type + composition] Motion: [camera movement type] Lighting: [lighting 设置up] Color: [palette + temperature] Material/Texture: [film grAIn / CGI / rea列出ic] Subject Action: [what the subject does]

Positive Prompt: [structured prompt: Subject + Action + 环境 + Camera + Lighting + Color Grade + Style + Technical] Negative Prompt: [what to exclude] Camera Prompt: [lens focal length, movement, angle] Style Prompt: [director reference, film stock, genre] Lighting Prompt: [specific lighting 设置up] Parameters: [aspect ratio, FPS, motion 扩展, CFG, 模型]

输出 格式化

Use this structure for every analysis:

`

Video Overall Style Analysis Video Type [Short film / Commercial / MV / Documentary / etc.] Overall Style [Atomic Punk + Post-apocalyptic Western / Cyberpunk / etc.] Director Reference [Most similar director(s) and why] Editing Rhythm [Slow build / Fast cuts / Montage / etc.] Estimated AI 模型(s) [Primary 模型 + supporting 工具s] AI Generation Artifacts 检测ed [列出 specific tells] Shot Breakdown Shot 01

[Full analysis per template above]

Shot 02

[Continue for ALL key shots]

Global Reverse-Engineered Prompts Global Style Prompt (应用lies to all shots)

[Master style anchor prompt]

Global Negative Prompt

[What to always exclude]

Global Camera Prompt

[Default lens, movement vocabulary, aspect ratio]

Parameter Estimation Parameter Value Aspect Ratio [e.g. 2.39:1] FPS [e.g. 24fps cinematic] Lens Range [e.g. 24mm-85mm] LUT Style [e.g. Bleach Bypass Warm] Color Temperature [e.g. 4500K warm] Depth of Field [e.g. Shallow f/1.4-2.8 for CU, Deep f/8-11 for WS] Shutter Feel [e.g. 180-degree shutter, 1/48s at 24fps] Film GrAIn [e.g. Medium, 35mm Tri-X 400 punch] Primary 模型 [e.g. 种子ance 2.0] Secondary 工具s [e.g. Midjourney for keyframes, DaVinci for grade] Reproduction 工作流 Keyframe Generation → Midjourney / Flux → 生成 character concept art + scene reference images Video Generation → Choose based on style: Ads / Product / Tech: H应用yHorse (快马) — Best for advertising & futuristic tech style Narrative / Short Film: Kling / 种子ance / 运行way → Text + reference image to video Creative / Art: Veo / Sora → Complex cinematic scenes Quick iteration: Pika / SVD → Short 命令行工具ps, rAPId 测试 Director Method → Write director-style prompts (WHY characters do things, not just WHAT) Audio 同步 → 种子ance immersive mode fo