360 Secure SkillHub
360 安全技能中心

infinimo-ai-design-product-image-create

通过 Infinimo AI Design 的产品图接口完成多步流程:查询目标平台与市场选项、获取模型字典、上传最多 12 张参考图、按平台与场景(主图/副图/详情 A+)提交生成任务并轮询结果,另可按 id 删除记录;需配置 Token,提示词与参考图至少提供其一。

anyunzhong
anyunzhong
热度 8.75
最近更新 2026-07-06 10:16:28
分类:创作依赖:第三方 API / 代码执行 / 读写文件应用场景:图像生成

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add infinimo-ai-design-product-image-create
Metadata
name
infinimo-ai-design-product-image-create
description
Generate e-commerce product images (hero, secondary, A+ detail) via Infinimo AI Design—marketplace/platform selection, model/aspect/resolution, reference uploads, submit and poll. Use for Amazon/Shopify listing heroes, lifestyle shots, and A+ modules.

Product Image Design

About Infinimo AI Design

Infinimo AI Design is an AI design platform for cross-border e-commerce product images and video. This skill calls the product image API (same as the web Product image design pages under /studio).

Web page: https://design.infinimo.ai/?source=q-i-d-clawhub

Studio tool paths (after sign-in):

Scene Path
Hero / main image /product-image-create?scene=cover
Secondary images /product-image-create?scene=cover_other
Detail / A+ /product-image-create?scene=detail

Authentication & base URL

Read INFINIMO_TOKEN or INFINIMO_API_KEY from the environment; never hardcode.

Common parameters

platform=1, terminal=4, language=en on all requests.


End-to-end flow

1. GET  /aigc/ec_product_media/platform_options   → target platforms & markets
2. GET  /aigc/ec_media/image/create/dic             → models / ratios / sizes
3. POST /upload/image                              → reference uploads (optional, max 12)
4. POST /aigc/ec_media/image/create                → submit product image job
5. GET  /aigc/ec_media/image/create/logs/product   → poll results

Poll until urls is non-empty or fail=true.

vs general image-create:

  • Required target_platform; optional region
  • Required image_scene (hero / secondary / detail)
  • Results via /logs/product, not /logs
  • Prompt or references: at least one required

1. Platform & market options

GET /aigc/ec_product_media/platform_options

bash scripts/platform_options.sh
Field Description
code Platform code → target_platform
name e.g. Amazon, Shopee
regions[].code Market code → region
regions[].name e.g. United States, Japan

2. Model dictionary

GET /aigc/ec_media/image/create/dic — same as general image generation.

bash scripts/dic.sh

3. Upload references

POST /upload/image — max 12 images, image/* only.

bash scripts/upload.sh /path/to/product.jpg

4. Submit product image generation

POST /aigc/ec_media/image/create

Parameter Required Description
target_platform yes Platform code from platform_options
image_scene no Default cover
region no Market code
prompt conditional At least prompt or images required
model, ratio, size no Dictionary ids
images no Reference URL array

image_scene values

Value Scene Studio alias
cover Hero / main ?scene=cover
cover_other Secondary ?scene=cover_other
detail Detail / A+ ?scene=detail

Reference tags in prompt: [Image 1], [Image 2], …

bash scripts/create.sh \
  --target-platform amazon \
  --scene cover \
  --region US \
  --prompt "White background hero, wireless earbuds, pro e-commerce photo" \
  --model MODEL_ID --ratio RATIO_ID --size SIZE_ID \
  --images '["https://cdn.example.com/ref.jpg"]'

5. Poll results

GET /aigc/ec_media/image/create/logs/product?start=1&size=5

bash scripts/logs.sh 1 5
items field Description
urls Output image URLs
fail true on failure
param Includes image_scene, target_platform, region

Poll every 3–5s until urls appear. Optional WebSocket: wss://www.clawec.com/api/aigc/socket → image_result_refresh.

Delete: GET /aigc/ec_media/image/log/delete?id=<id>

See references/response-schema.md.


Workflow

  1. Confirm scene (hero/secondary/detail), platform, and market
  2. Ensure token is set
  3. platform_options → pick target_platform / region
  4. dic → model / ratio / size
  5. upload references if needed
  6. create (prompt and/or images)
  7. Poll logs/product and return URLs

Example

Input: Amazon US hero image for wireless earbuds using uploaded product photo

  1. platform_options.sh → amazon, US
  2. dic.sh → model / ratio / size
  3. upload.sh product.jpg
  4. create.sh --target-platform amazon --region US --scene cover --prompt "White hero, wireless earbuds" --images '[...]'
  5. logs.sh → output URL