fashion-mannequin-three-view

name	fashion-mannequin-three-view
description	Turn uploaded fashion product reference images into a consistent 16:9 white-background three-view apparel product sheet on identical headless armless mannequins. Use when the user provides front/back/side garment references and asks for clothing 三视图, 正面/侧面/背面 views, product consistency fixes, ecommerce apparel view sheets, or GPT Image 2 prompts/generation for garment multi-view displays.

Fashion Mannequin Three View

Workflow

Use this skill to make a prompt, or to generate the image directly when the user asks to generate.

1. Inspect every uploaded product reference and extract one shared garment specification before writing any prompt:
   • garment type, length, silhouette, fit, fabric weight, color
   • neckline, straps/sleeves, closures, ties, bows, waist/hip shaping
   • seams, ruching, drape panels, pleats, slits, hem shape
   • front-only, back-only, and ambiguous details
2. Treat the shared garment specification as the source of truth for all views. Do not describe each angle as a separate outfit.
3. If a side reference is missing, infer the side view conservatively from the front and back construction. Say that the side is derived from the shared garment structure, not invented as a new design.
4. Use identical mannequins in all three panels:
   • featureless fashion dress-form / mannequin
   • headless, armless, no hands, no face, no hair
   • matte white or very light gray surface
   • same height, same body proportions, same scale, same neutral upright posture
5. Use a clean product-spec layout:
   • 16:9 horizontal canvas
   • pure white background #FFFFFF
   • three evenly spaced full-length views, left to right: front, side, back
   • centered Chinese labels below each view: 正面视图, 侧面视图, 背面视图
6. When the user asks for prompt engineering only, return the final GPT Image 2 prompt. When the user says generate/make the image, call image generation with that prompt.

Prompt Rules

Keep the prompt explicit about consistency. Repeat that all three views show the same garment on the same mannequin type.

Prefer product-render language over fashion-editorial language:

• use: orthographic product sheet, apparel technical presentation, neutral ecommerce lighting
• avoid: lifestyle, runway, garden, beautiful model, cinematic, editorial pose

Do not include props, environments, jewelry, hands, drinks, chairs, tables, shadows from scenery, or realistic human models unless the user explicitly asks.

Do not ask the model to “design a similar dress.” Ask it to “reconstruct the same garment from the references.”

Consistency Checklist

Before finalizing the prompt or generation request, verify the prompt locks these items:

• Same color and fabric in all views
• Same garment length and hem width in all views
• Same neckline/strap construction from front to side to back
• Same waist/hip ruching or drape placement where structurally visible
• Back view preserves back-only details such as ties, bow, open back, zipper, buttons, or closure
• Side view is a neutral side profile, not a new angled pose
• Labels are exactly the requested Chinese strings, with no extra text

Reference

For a reusable GPT Image 2 prompt template, load references/gpt-image-2-template.md.