---
name: ecommerce-ai-model-images
description: 'Generate ecommerce AI model images from product photos and model requirements. Use when the user asks to create clothing try-on style images, bag carrying images, accessory wearing images, beauty usage images, or marketplace-ready AI model visuals with GPT Image 2.'
metadata:
  icon: "\U0001F9CD"
  title: Ecommerce AI Model Images
  companionSkill: gpt-image-2
  packageName: '@zjxuyunshi/ecommerce-ai-model-images'
  version: 0.0.1
---

# Ecommerce AI Model Images

## When to Use

Use this skill when the user wants to create ecommerce model images from product photos without relying on an existing model photo. Typical requests include:

- Clothing try-on style images, garment-on-model images, robe/dress/top/pants model shots.
- Bag hand-carrying, shoulder-carrying, crossbody, or outfit-match images.
- Accessories wearing images, shoes-on-foot images, beauty/personal-care usage images.
- AI model shots for Taobao, Tmall, 1688, Amazon, Shopee, Temu, Xiaohongshu, Douyin, or social commerce.
- Generate model images for a target audience, such as vacation style, commuter style, young female, mature female,欧美,日韩, Southeast Asia, or minimalist studio.

Use this when the main job is to create a new model shot. Use `ecommerce-model-replace-images` when the user supplied a model/reference scene to replace. Use `ecommerce-pose-variant-images` for multiple pose variations of a same model/product setup.

## Inputs

- Required: original product image. Preserve product color, print, shape, logo placement, pattern details, garment silhouette, fabric drape, accessory details, and distinctive features.
- Optional but useful: product category and wearing/use method, such as `罩袍上身`, `手提包`, `斜挎包`, `耳环佩戴`, `鞋子上脚`, or `护肤使用`.
- Optional: model requirements: gender, age range, region/aesthetic, body type, styling, hair, pose, skin tone, and expression. Keep requests respectful and commercially appropriate.
- Optional: platform and scene: white studio, premium studio, beach vacation, street snap, commuting, home, outdoor, Xiaohongshu note cover, Amazon clean model shot, etc.
- Optional: reference images for product fidelity, pose, mood, styling, or scene.
- Optional: output count, size, and format.

Do not invent additional product variants, colors, sizes, bundles, claims, or logos. Do not create sexualized or exploitative model imagery.

## Execution

Use `$gpt-image-2` through the bundled edit script. Prefer:

```bash
python3 "/Users/yunshi/.codex/skills/gpt-image-2/scripts/edit_image.py" \
  --prompt-file "<prompt.txt>" \
  --image "<product-image>" \
  --image "<optional-reference-image>" \
  --out-dir "<output-dir>" \
  --name "ai-model-01" \
  --output-format "png" \
  --quality "high" \
  --size "1024x1536"
```

Prompt structure:

```text
Create a marketplace-ready ecommerce AI model image.
Product: preserve image 1 accurately, including color, print placement, shape, silhouette, fabric drape, logos, labels, and distinctive product details.
Model direction: <target model gender/age/aesthetic/body/styling if supplied, otherwise choose a platform-safe commercial model>.
Wearing/use method: <how the product should be worn/carried/used>.
Scene/platform: <studio/lifestyle/platform style>.
Reference usage: use images 2+ only for pose, mood, scene, product fidelity, or styling inspiration.
Composition: product is clearly visible, model pose is natural, anatomy and hands are realistic, fit/scale is believable, clean ecommerce lighting.
Avoid: changed product print/color, hidden product, unrealistic body, distorted hands/face, celebrity/private-person likeness, extra products, fake logos, watermarks, QR codes, unsupported claims.
```

## Result Handling

The script returns JSON. Read `local_paths`, `remote_urls`, `uploads`, and `metadata`. Save outputs under `/Users/yunshi/Downloads/ecommerce-ai-model-images/<short-product-name>-<timestamp>/` and preview the first image when practical.

## Failure Handling

- Missing product image: ask for one.
- Missing category/use method: infer conservative defaults only when visible; otherwise ask.
- Product fidelity failure: retry with product image first and stronger preservation language.
- Unsafe model request: refuse or redirect to respectful, platform-safe commercial imagery.
- Script failure: report the JSON error, stderr progress, and smallest next fix.