| name | dreamy-motion-editorial |
|---|---|
| description | Extract the semantic hero from any casual photo and rebuild it as a premium hazy dreamlike motion-editorial image with large negative space, simplified context, selective preservation, and restrained poster hierarchy. Use when the user provides or references a photo and asks for 高级感、朦胧感、梦幻感、电影感、慢门拖影、柔焦、胶片感、诗意海报、Pinterest 同类氛围、图生图提示词、AI 生图提示词、照片风格改造、帖子视觉、封面图、艺术摄影、青春残影、玻璃倒影、花卉虚焦、逆光晕染, or wants a reusable photo-to-poster art-direction workflow. Produces an adaptive diagnosis, mechanism selection, Chinese and English image-to-image prompts, negative prompt, preservation constraints, strength and ratio guidance, and optional restrained poster copy. |
Dreamy Motion Editorial
Turn the supplied photograph into an art-directed photographic event, not a faithful full-scene reconstruction or generic blur filter. Preserve the semantic hero and essential action, then simplify, fade, remove, crop, or reposition nonessential content to create editorial hierarchy.
Default Composition Policy
Default to semantic-core, not full-scene fidelity.
- Use Keep / Fade / Remove before writing prompts.
- Keep one hero, one essential gesture or object, and no more than two contextual cues.
- Use kinetic negative space, not blank empty space: allocate 35–55% of the frame to broad directional motion veils, 15–30% to quiet tonal fields, and the remainder to the sharp hero.
- Let the hero occupy roughly 25–40% of the frame unless the source requires a close portrait.
- Permit reframing, asymmetric placement, subject repositioning, and background reconstruction.
- Fade by transformation, not erasure: stretch 2–4 secondary people, flowers, fabric, vehicles, paper, foliage, or architectural edges into large directional blur masses, translucent streaks, and light-color trails.
- Remove only tiny clutter and competing details. Keep enough secondary forms to build foreground, midground, and background motion depth.
- Preserve the entire scene only when the user explicitly requests
identity-lockor documentary fidelity.
The output should feel extracted, kinetic, and composed, not merely color-graded. A successful result has fewer readable objects than the source but more large-scale motion energy: a sharp anchor surrounded by moving translucent forms.
Workflow
Inspect the photo visually. Identify the hero, identity-sensitive details, pose, relationships, clothing, environment, existing light direction, depth layers, reflective surfaces, plausible motion, native colors, and structural risks.
Extract the semantic core. Write three short lists: Keep, Fade, and Remove. Keep only the hero identity, essential gesture or object, one native accent color, and no more than two context cues.
Design kinetic negative space. Specify three depth layers: near-lens directional motion veil, midground secondary-form drag, and background light/color trails. Empty quiet fields are supporting space only, never the main effect.
Select the route. Read
references/style-system.md. Choose exactly one primary mechanism and at most one support mechanism:crowd-driftfloral-diffusionglass-memorybacklit-halationlow-angle-color-fieldpassing-gesture
Use the helper when useful. Convert observed traits into a route recommendation with:
python scripts/analyze_photo.py --subject portrait --light backlit --motion walking --depth layered --background street --risks face --strength dream-haze --json
Treat the result as a consistency aid, not a replacement for visual judgment. The script analyzes supplied observations; it does not inspect pixels.
Choose strength. Default to
dream-haze. Userestrained-premiumfor brands, products, identity-sensitive portraits, or subtle requests. Useexperimental-memoryonly when the user clearly wants stronger abstraction.Build the prompt package. Read
references/prompt-recipes.md. Follow the required order:source preservation → composition → action/motion → lighting → color → optical behavior → analog texture → emotional tone → rendering constraintsAdd negative controls. Include anatomy, identity, blur, optical, color, and cheap-aesthetic exclusions relevant to the actual photo. Do not dump irrelevant negative keywords.
Add poster direction only when requested. Keep typography under 12% of the frame, use existing negative space, and suggest adding exact long text in a design tool rather than relying on image-generation spelling.
Verify before delivery. Apply
references/quality-rubric.md. Rewrite if preservation, sharpness hierarchy, or physical motivation scores zero.Generate only when explicitly requested. Default to prompt output. When the user asks to create or edit the actual raster image, invoke the available image-generation/image-editing tool and use the finished prompt package as the generation brief.
Route Selection Rules
- Prefer a feature already present in the source over an invented decorative element.
- Assign motion and diffusion mainly to secondary layers, and make the selected motion mechanism visibly occupy 35–55% of the default dream-haze frame.
- Keep one face plane, eye line, gesture, product edge, label, silhouette, or flower center readable.
- Use flowers or petals only when present, semantically plausible, or explicitly requested.
- Use glass or reflection only when it improves depth and can be physically explained.
- Do not choose
crowd-driftfor isolated products or quiet selfies without plausible surrounding movement. - Retain 2–4 large secondary motion forms when the source supports people, flowers, traffic, fabric, paper, foliage, or nearby objects.
- Do not combine more than two mechanisms.
- Do not imitate a named living artist or a single Pinterest creator. Translate transferable visual principles instead.
Strength Defaults
Restrained Premium
- Preserve the hero while allowing subtle movement across 25–40% of the frame.
- Use one subtle optical cue.
- Keep bloom, grain, and color separation restrained.
- Default for products, brand work, and recognition-sensitive portraits.
Dream Haze
- Preserve the hero while using 35–55% of the frame for visible directional motion veils and trails.
- Make the chosen motion, reflection, or diffusion clearly visible.
- Allow one compatible support mechanism.
- Default for social posts and editorial portraits.
Experimental Memory
- Use stronger motion drag, doubling, refraction, or color contamination.
- Preserve one unmistakable visual anchor.
- Do not allow semantic collapse, face replacement, or arbitrary scene invention.
Output Format
Deliver in this structure:
## 照片诊断
- 主体：
- 必须保留：
- 可利用的光线/动态/前景：
- 风险：
## 构图提炼
- Keep：
- Fade：
- Remove：
- 动态覆盖：
- 安静留白：
- 前景掠影：
- 中景拖影：
- 背景光色流：
- 主体占比与位置：
## 视觉方案
- 主机制：
- 辅助机制：无 / ...
- 强度：克制高级 / 梦幻朦胧 / 强烈实验
- 选择理由：
## 中文图生图提示词
...
## English Image-to-Image Prompt
...
## 负面提示词
...
## 保真约束
- ...
## 建议参数
- 比例：
- 重绘强度：
- 主体保真优先级：
## 备选方向（可选）
...
## 海报文案与排版（仅在需要时）
...
Parameter Guidance
Use model-neutral language unless the user names a generator.
- Portrait post:
4:5 - Editorial poster:
3:4 - Story or vertical cover:
9:16 - Environmental scene: preserve the source ratio or use
4:5 - Recognition-sensitive image-to-image: low-to-medium denoise or transformation strength
- Dream Haze: medium transformation strength
- Experimental Memory: medium-high transformation strength while locking identity or reference strength when supported
Never invent unsupported parameter names. If the user names a model, adapt terminology to that model's current interface.
Failure Modes
Reject and rewrite prompts that:
- preserve every background object with equal clarity
- produce a literal full-scene reconstruction with no hierarchy
- produce a static empty background with no directional motion atmosphere
- erase every secondary person or object instead of transforming some into motion masses
- keep more than two contextual cues by default
- blur the whole frame uniformly
- replace physical optical causes with abstract “dream magic”
- add random flowers, sparkles, jewelry, costumes, or people
- beautify the subject into a generic face
- request white fog over the entire image
- stack every mechanism into one scene
- combine sterile 8K HDR sharpness with analog softness
- use extreme teal-orange grading
- turn the result into wedding-studio, cosplay, cyberpunk, or stock-photo imagery
- place dense text over faces, gestures, or product labels
Resource Map
- Read
references/style-system.mdfor diagnosis, route grammar, color, texture, and strength logic. - Read
references/prompt-recipes.mdfor bilingual templates, mechanism modules, preservation add-ons, negative prompts, and examples. - Read
references/quality-rubric.mdfor scoring and repairs. - Run
scripts/analyze_photo.pywhen a structured route recommendation improves consistency.