Ecommerce Scene Images

When to Use

Use this skill when the user wants to place a real product into a believable ecommerce scene: home, office, bedroom, living room, kitchen, bathroom, outdoor, travel, commute, studio, seasonal, gift, or retail display.

Use this for scene renderings and lifestyle listing images. Use ecommerce-use-case-images when the priority is explaining a specific usage scenario or target user, and ecommerce-background-replace-images when the request is mostly background cleanup or replacement.

Inputs

Required: one or more product images. Preserve product identity, shape, color, scale cues, logos, packaging, and distinctive features.
Optional: scene requirement, such as 客厅, 通勤, 旅行, 梳妆台, 户外露营, 节日礼品, or 高级棚拍. If absent, choose a clean platform-safe commercial scene for the category.
Optional: reference images for scene mood, lighting, props, camera angle, or composition. Do not copy competitor products, watermarks, logos, or text.
Optional: platform, target language, text policy, output size, format, quality, and count. Default to no visible text unless requested.
Optional: allowed props. Use generic, category-relevant props only; do not add extra products that imply an included bundle.

Do not invent product claims. Keep the product naturally integrated but clearly dominant.

Execution

Use $gpt-image-2 image editing mode through fusion-api.openai_image_edit_async_submit. Upload local files with oo file upload, pass downloadUrl values in images, product images first and scene references after.

Prompt structure:

Create a realistic marketplace-ready ecommerce scene image.
Primary product: preserve image 1 accurately, including shape, proportions, color, material appearance, logos, labels, and distinctive details.
Scene: <requested scene or category-safe default>.
Reference usage: use images 2+ only for mood, lighting, props, perspective, and composition; do not copy unrelated products, brands, text, or watermarks.
Composition: product is the hero, naturally placed, realistic scale, clean commercial lighting, uncluttered background, platform-safe visual style.
Text policy: <no text by default, or minimal factual text if requested>.
Avoid: distorted product, hidden product, fake usage claims, extra bundled items, QR codes, phone numbers, social handles, platform badges, watermarks, competitor marks, messy props.

Run:

oo connector run "fusion-api" --action "openai_image_edit_async_submit" --data @payload.json --json

Use model, prompt, images, output_format, quality, size, and optional n exactly as defined by $gpt-image-2.

Result Handling

Read image URLs from .data.data[].url. Download to /Users/yunshi/Downloads/ecommerce-scene-images/<short-product-name>-<timestamp>/ with names like scene-image-01.png. Preview the saved image when practical and note the selected scene.

Failure Handling

Missing product image: ask for one.
Scene is ambiguous: choose a clean category-safe scene or ask one focused question if the scene changes the business outcome.
Product becomes too small or altered: retry with stronger preservation and product-dominance wording.
Unsafe or proof-dependent usage claims: omit them.
Connector handle, timeout, auth, billing, upload, or schema failure: follow $gpt-image-2 guidance and report the exact blocker.