Jimeng AI Generation Skill

Overview

• 🎨 Text-to-Image: Generate high-quality images from text descriptions
• 🎭 Image Synthesis: Intelligently merge and blend multiple images
• 🎬 Text-to-Video: Create short videos from text prompts
• 🎞️ Image-to-Video: Add animation effects to static images

• Users request to generate, create or produce images or videos
• Users mention "jimeng", "Jimeng" or request AI visual content generation
• Users provide text descriptions and expect visual outputs
• Users want to combine, merge or synthesize multiple images
• Users want to add animation or motion effects to static images

Prerequisites

1. Server Must Be Running, in one of the following modes:
   • stdio Mode: Configured in MCP clients (Claude Desktop, Cherry Studio)
   • SSE Mode: Run as an HTTP server with SSE transmission
   • HTTP Mode: Run as a REST API server
2. Environment Variables Configured:

   • JIMENG_API_KEY

     : Your Jimeng API key (obtained from Jimeng website cookies)

   • JIMENG_API_URL

     : API endpoint (default: http://127.0.0.1:8001)

   • JIMENG_MODEL

     : Model name (default: jimeng-4.5)
3. Backend API Running: The jimeng-free-api-all Docker container must be active

references/setup_guide.md

Quick Start

Basic Usage Workflow

1. Identify Task Type based on user input
2. Extract Required Parameters from the request
3. Call the Corresponding jimeng-mcp-server Tool
4. Return Generated Content URLs to the user

Example Requests

User: "Generate an image with Jimeng: Shiba Inu under cherry blossom trees"
→ Use the text_to_image tool with parameter prompt="Shiba Inu under cherry blossom trees"

User: "Help me synthesize these two images, with the style leaning towards the first one"
→ Use the image_composition tool and provide image URLs

User: "Create a 5-second video: Scene of a pony crossing a river"
→ Use the text_to_video tool, set the prompt and duration

User: "Add animation effects to this image"
→ Use the image_to_video tool and provide the image URL

Core Capabilities

1. Text-to-Image

text_to_image

• prompt

  (required): Text description of the desired image

• model

  (optional): Model version (default: jimeng-4.5)

• ratio

  (optional): Image aspect ratio ("1:1", "4:3", "3:4", "16:9", "9:16")

• resolution

  (optional): Resolution preset ("1k", "2k", "4k", default: 2k)

• negativePrompt

  (optional): Elements to avoid in the generated image

• 16:9 → Landscape/widescreen (video covers, banners)
• 1:1 → Square (avatars, social media)
• 9:16 → Portrait/mobile screen (short video covers)
• 4:3 → Standard landscape (blog illustrations)
• 3:4 → Standard portrait (portrait photos)

# User request: "Generate an image: Beach at sunset with coconut trees"
{
  "model": "jimeng-4.5",
  "prompt": "Beach at sunset with coconut trees",
  "ratio": "16:9",
  "resolution": "2k"
}

• Higher resolution (4k) is suitable for print and high-quality displays
• Lower resolution (1k) is suitable for quick previews
• Use descriptive prompts for better results
• Specify art style, lighting, and atmosphere to enhance control

2. Image Synthesis

image_composition

• prompt

  (required): Description of how to synthesize the images

• images

  (required): Array of 2-5 image URLs to synthesize

• model

  (optional): Model version (default: jimeng-4.5)

• ratio

  (optional): Output image aspect ratio ("1:1", "4:3", "3:4", "16:9", "9:16")

• resolution

  (optional): Resolution preset ("1k", "2k", "4k", default: 2k)

# User request: "Synthesize these two images, retaining the style of the first one"
{
  "model": "jimeng-4.5",
  "prompt": "Seamlessly blend the two images while maintaining the artistic style of the first image",
  "images": [
    "https://example.com/image1.jpg",
    "https://example.com/image2.jpg"
  ],
  "ratio": "4:3",
  "resolution": "2k"
}

• Blend portraits with backgrounds
• Style transfer between images
• Create artistic composite works
• Merge elements from multiple photos

• Provide clear synthesis instructions in the prompt
• Images should have compatible resolutions
• Describe the desired blending style (seamless, artistic, realistic)

3. Text-to-Video

text_to_video

• prompt

  (required): Text description of the video scene

• model

  (optional): Model version (default: jimeng-video-3.0)

• ratio

  (optional): Video aspect ratio ("16:9", "9:16", "4:3", "3:4", "1:1")

• resolution

  (optional): Preset resolution ("480p", "720p", "1080p")

• "480p" → Quick preview
• "720p" → Balanced quality/speed (recommended)
• "1080p" → High quality

# User request: "Generate a 5-second video: Kitten fishing"
{
  "model": "jimeng-video-3.0",
  "prompt": "An orange kitten sitting by the river, holding a fishing rod and focusing on fishing, sunny weather",
  "ratio": "16:9",
  "resolution": "720p"
}

• Duration: Typically 3-5 seconds
• Format: MP4
• Generation Time: 30-60 seconds
• Frame Rate: 24-30 fps

• Include scene details, actions, and atmosphere
• Keep prompts focused on a single clear action
• Specify time of day, weather, or mood for better results
• Start with 720p to balance quality and speed

4. Image-to-Video Animation

image_to_video

• prompt

  (required): Description of the desired animation effect

• file_paths

  (required): Array of image URLs to animate

• model

  (optional): Model version (default: jimeng-video-3.0)

• ratio

  (optional): Video aspect ratio ("16:9", "9:16", "4:3", "3:4", "1:1")

• resolution

  (optional): Preset resolution ("480p", "720p", "1080p")

# User request: "Animate this photo with gentle camera zoom"
{
  "model": "jimeng-video-3.0",
  "prompt": "Add gentle motion effects and natural camera zoom to create a cinematic feel",
  "file_paths": ["https://example.com/photo.jpg"],
  "ratio": "16:9",
  "resolution": "720p"
}

• Character motion
• Camera movements
• Scene transitions
• Environmental effects (wind, rain, etc.)

• Describe the desired type of motion
• Consider image content when selecting effects
• Portrait photos suit subtle movements
• Landscape photos suit pan/zoom effects

Workflow Guide

Decision Tree

Receive User Request
    │
    ├─ Contains "generate image" or "create image"?
    │   └─ Yes → Use text_to_image
    │
    ├─ Contains "synthesize" or "merge/blend images"?
    │   └─ Yes → Use image_composition
    │
    ├─ Contains "generate video" or "create video"?
    │   └─ Yes → Use text_to_video
    │
    └─ Contains "animate" or "animate image"?
        └─ Yes → Use image_to_video

Parameter Extraction

1. Extract Prompt: User's description of the desired content
2. Identify Aspect Ratio: Extract size preferences (landscape/portrait/square) corresponding to the ratio parameter
3. Parse Resolution Requirements: Look for quality requirements corresponding to the resolution parameter
4. Collect Image URLs: For synthesis and animation tasks

Error Handling

1. Check Server Status: Verify if jimeng-mcp-server is running
2. Validate API Key: Ensure JIMENG_API_KEY is configured
3. Check Parameters: Confirm all required fields are provided
4. Check Image URLs: Verify URLs for synthesis/animation are accessible
5. Report Errors Clearly: Explain the problem and suggest solutions

• API key not configured

  : Set JIMENG_API_KEY in the environment

• Server not responding

  : Start the jimeng-free-api-all Docker container

• Invalid image URL

  : Ensure the URL is publicly accessible

• Generation timeout

  : Large videos may take 60+ seconds

Advanced Usage

Combine Multiple Tools

1. Use

   text_to_image

   to generate a base image
2. Use

   image_to_video

   to add animation to the result

1. Use

   image_composition

   to synthesize images
2. Generate variants with adjusted prompts

Optimization Tips

• Use lower resolution (720p instead of 1080p, or 1k instead of 2k)
• Keep prompts concise yet descriptive

• Use detailed, specific prompts
• Select appropriate ratio based on the scene
• Use higher resolution (2k or 4k)
• Specify art style and techniques
• Include lighting and atmosphere descriptions

Batch Processing

1. Process requests sequentially (one at a time)
2. Provide progress updates for each item
3. Collect all results before final response
4. Consider resource limits (API quotas)

Troubleshooting

Server Connection Issues

1. Check if the jimeng-free-api-all Docker container is running:
   bash

   docker ps | grep jimeng

2. Verify server accessibility:
   bash

   curl http://127.0.0.1:8001/health

3. Restart the Docker container if needed

API Key Issues

1. Verify JIMENG_API_KEY in the .env file
2. Obtain a new API key from Jimeng website cookies (sessionid value)
3. Ensure the key format is correct (no extra spaces or quotes)

Generation Quality Issues

1. Optimize prompts with more specific details
2. Adjust the

   ratio

   parameter to select an appropriate aspect ratio
3. Try different

   resolution

   settings
4. Add

   negativePrompt

   to exclude unwanted elements

Timeout Errors

1. Video generation typically takes 30-60 seconds - please be patient
2. If timeouts persist, try lower resolution
3. Check server resource usage
4. Verify network connection to Jimeng API

Resources

references/

• setup_guide.md

  : Detailed installation and configuration instructions

• api_reference.md

  : Complete API documentation for all tools

Project Links

• GitHub Repository: https://github.com/wwwzhouhui/jimeng-mcp-server
• Backend API: https://github.com/wwwzhouhui/jimeng-free-api-all
• Jimeng Official Website: https://jimeng.jianying.com/

Best Practices

1. Always Verify Server Status Before Attempting Generation
2. Use Appropriate Resolution Based on Use Case and Speed Requirements (ratio controls aspect ratio, resolution controls clarity)
3. Provide Detailed Prompts for Better Generation Quality
4. Handle Errors Gracefully and Provide Clear User Feedback
5. Consider Rate Limits When Processing Multiple Requests
6. Test with Simple Prompts Before Complex Synthesis
7. Cache Frequently Used Parameters such as preferred ratio and resolution

Limitations

• Free Tier Limits: Official Jimeng API allows 66 credits per day
• Video Duration: Typically limited to 3-10 seconds
• Generation Time: Videos may take 30-60 seconds to generate
• Image Synthesis: Best results with 2-3 images, maximum 5 images supported
• Server Dependency: Requires jimeng-free-api-all backend to run
• Network Requirements: Internet access required to call Jimeng API