Ecommerce Background Replace Images
When to Use
Use this skill when the user wants to keep the product but replace, clean, or upgrade the background. Typical requests include white background, transparent-feel background, premium studio background, catalog consistency, seasonal background, light lifestyle background, or removing messy source environments.
Use this when the product itself should remain essentially unchanged. Use ecommerce-main-images when the whole cover composition is being redesigned and ecommerce-scene-images when the request is a richer lifestyle scene.
- Required: product image. Preserve product shape, color, proportions, logos, packaging, labels, and distinctive details.
- Optional: desired background type:
white background, premium studio, light lifestyle, seasonal, category scene, brand color, or catalog unified.
- Optional: reference images for background style, lighting, color palette, surface, and composition.
- Optional: platform, target ratio, output size, format, quality, and count.
- Optional: shadow/reflection preference. Default to natural soft shadow.
Do not alter the product unless the user explicitly asks for product retouching.
Execution
Use $gpt-image-2 image editing mode through fusion-api.openai_image_edit_async_submit. Upload the product image first, then optional background references.
Prompt structure:
Create an ecommerce background replacement image.
Primary product: preserve image 1 accurately, including shape, color, proportions, material appearance, logos, labels, packaging, and distinctive details.
Background goal: <white background | premium studio | light lifestyle | seasonal | brand color | catalog unified>.
Reference usage: use images 2+ only for background style, lighting, surface, and color palette; do not copy unrelated products, text, logos, or watermarks.
Composition: product remains the hero, clean edges, realistic contact shadow/reflection, platform-ready crop, uncluttered background.
Avoid: product distortion, changed colors, removed logos, fake props implying included items, QR codes, phone numbers, platform badges, watermarks, unsupported claims.
Run:
oo connector run "fusion-api" --action "openai_image_edit_async_submit" --data @payload.json --json
Use exact $gpt-image-2 payload fields.
Result Handling
Read URLs from .data.data[].url. Download to /Users/yunshi/Downloads/ecommerce-background-replace-images/<short-product-name>-<timestamp>/ with names like background-replace-01.png. Preview when practical and mention the selected background goal.
Failure Handling
- Missing product image: ask for one.
- Ambiguous background request: default to premium studio, or ask if white-background compliance matters.
- Product changes too much: retry with stronger preservation wording.
- Too many references: choose the clearest background/style references.
- Connector handle, timeout, auth, billing, upload, or schema failure: follow
$gpt-image-2 guidance and report the exact blocker.
---
name: ecommerce-background-replace-images
description: 'Replace or upgrade product photo backgrounds for ecommerce images. Use when the user asks to change a rough background into white background, premium studio, light lifestyle, seasonal, platform-ready, or consistent catalog backgrounds while preserving the original product with GPT Image 2.'
metadata:
icon: "\U0001FA84"
title: Ecommerce Background Replace Images
companionSkill: gpt-image-2
packageName: '@zjxuyunshi/ecommerce-background-replace-images'
version: 0.0.1
---
# Ecommerce Background Replace Images
## When to Use
Use this skill when the user wants to keep the product but replace, clean, or upgrade the background. Typical requests include white background, transparent-feel background, premium studio background, catalog consistency, seasonal background, light lifestyle background, or removing messy source environments.
Use this when the product itself should remain essentially unchanged. Use `ecommerce-main-images` when the whole cover composition is being redesigned and `ecommerce-scene-images` when the request is a richer lifestyle scene.
## Inputs
- Required: product image. Preserve product shape, color, proportions, logos, packaging, labels, and distinctive details.
- Optional: desired background type: `white background`, `premium studio`, `light lifestyle`, `seasonal`, `category scene`, `brand color`, or `catalog unified`.
- Optional: reference images for background style, lighting, color palette, surface, and composition.
- Optional: platform, target ratio, output size, format, quality, and count.
- Optional: shadow/reflection preference. Default to natural soft shadow.
Do not alter the product unless the user explicitly asks for product retouching.
## Execution
Use `$gpt-image-2` image editing mode through `fusion-api.openai_image_edit_async_submit`. Upload the product image first, then optional background references.
Prompt structure:
```text
Create an ecommerce background replacement image.
Primary product: preserve image 1 accurately, including shape, color, proportions, material appearance, logos, labels, packaging, and distinctive details.
Background goal: <white background | premium studio | light lifestyle | seasonal | brand color | catalog unified>.
Reference usage: use images 2+ only for background style, lighting, surface, and color palette; do not copy unrelated products, text, logos, or watermarks.
Composition: product remains the hero, clean edges, realistic contact shadow/reflection, platform-ready crop, uncluttered background.
Avoid: product distortion, changed colors, removed logos, fake props implying included items, QR codes, phone numbers, platform badges, watermarks, unsupported claims.
```
Run:
```bash
oo connector run "fusion-api" --action "openai_image_edit_async_submit" --data @payload.json --json
```
Use exact `$gpt-image-2` payload fields.
## Result Handling
Read URLs from `.data.data[].url`. Download to `/Users/yunshi/Downloads/ecommerce-background-replace-images/<short-product-name>-<timestamp>/` with names like `background-replace-01.png`. Preview when practical and mention the selected background goal.
## Failure Handling
- Missing product image: ask for one.
- Ambiguous background request: default to premium studio, or ask if white-background compliance matters.
- Product changes too much: retry with stronger preservation wording.
- Too many references: choose the clearest background/style references.
- Connector handle, timeout, auth, billing, upload, or schema failure: follow `$gpt-image-2` guidance and report the exact blocker.