image-creator-external（外部图像创建者）

Image Creator

Generate images using ShortArt backend API, supporting text-to-image and image-to-image.

Authentication Setup (Required for First Use)

image-creator requires a ShortArt API Key. There are two ways to configure it:

Method 1: OpenClaw Config (Recommended)

Add it to ~/.openclaw/openclaw.json:

{
  "skills": {
    "entries": {
      "image-creator-external": {
        "env": { "SHORTART_API_KEY": "paste_your_api_key_here" }
      }
    }
  }
}

Method 2: Shell Environment Variable

Add to ~/.zshrc (or ~/.bashrc for bash):

export SHORTART_API_KEY="paste_your_api_key_here"

Then reload: source ~/.zshrc

You can obtain the API Key from the ShortArt console. Note for LLM: When executing scripts, if API key is not found, try running with source ~/.zshrc && python3 scripts/impl.py... to load environment variables from shell config.

Parameter Quick Reference

Execution Method

IMPORTANT: Always prefix commands with source ~/.zshrc && to load environment variables.

Generate images by calling scripts/impl.py:

source ~/.zshrc && python3 scripts/impl.py "<prompt>" \
  --model {ModelName} \
  --count {Count} \
  --resolution {Resolution} \
  --aspect-ratio {AspectRatio}

Image-to-image (with reference image):

# Use existing OSS path
source ~/.zshrc && python3 scripts/impl.py "<prompt>" --image "images/20260121/.../filename.jpg"

# Upload local image
source ~/.zshrc && python3 scripts/impl.py "<prompt>" --upload /path/to/local/image.jpg

Return Result

After generation completes, the script returns:

{
  "status": "success",
  "project_id": "{projectId}",
  "credit": "{credit}",
  "sub_credit": "{subCredit}",
  "consumed_credit": "{consumedCredit}",
  "images": [
    {
      "id": "{imageId}",
      "path": "images/20260314/xxx/filename.jpg",
      "url": "https://file.shortart.ai/images/20260314/xxx/filename.jpg",
      "width": 1024,
      "height": 1024
    }
  ],
  "domain": "https://file.shortart.ai"
}

Workflow

1. Understand Requirements — Confirm subject, style, dimensions, quantity
2. Optimize Prompt — Refer to references/prompt-guide.md to expand description
3. Select Parameters — Choose model / resolution / aspectRatio based on use case
4. Submit Task — Run scripts/impl.py to submit the generation task: source ~/.zshrc && python3 scripts/impl.py "<prompt>" --model {model} --count {count} This returns immediately with project_id
5. Ask User to Wait — After task submission, ask user: "Task submitted successfully (project_id: xxx). Do you want to wait for the generation to complete?"
6. Poll for Completion (if user agrees) — Run polling command: source ~/.zshrc && python3 scripts/impl.py --poll {project_id} --model {model} --count {count} --resolution {resolution}
7. Handle Result — Based on the result status:

- Success: Ask user if they want to download images - Timeout/Failed: Ask user if they want to retry

1. Download Images (if user agrees) — Run download command with the full JSON result: source ~/.zshrc && python3 scripts/impl.py --download '{json_result}' Images will be saved to ~/Downloads/ with filenames like shortart_20260315_143022_1.jpg
2. Inform User — Tell user the local file paths where images were saved

IMPORTANT: Never expose OSS URLs directly to users. Always download images locally first.

Polling Parameters

The script estimates completion time based on:

• nano-banana-pro: ~40s, polls every 5-6s
• nano-banana-2: ~43s, polls every 6s
• seedream4.5: ~74s, polls every 8s
• Timeout: 2-5 minutes depending on parameters

Reference Files

• API Parameters & Response Format → references/api.md
• Prompt Writing Templates → references/prompt-guide.md