johnnyphp/mockofun-omnigen repository overview

⁠MockoFun OmniGen Image

Official Docker image for MockoFun's advanced image generation service, powered by OmniGen. This image provides the serverless backend supporting MockoFun's⁠ AI-powered image generation capabilities.

⁠Features

• Text-to-image generation
• Image-to-image generation with multi-image input support
• Pre-loaded OmniGen model for faster startup
• Runpod serverless compatibility
• CUDA support for GPU acceleration

⁠Usage

This image is designed to run on Runpod's serverless infrastructure. The handler accepts two types of requests:

⁠Text to Image Generation

{
    "input": {
        "prompt": "A detailed description of the desired image",
        "height": 1024,        // optional, default: 1024
        "width": 1024,         // optional, default: 1024
        "guidance_scale": 2.5, // optional, default: 2.5
        "seed": 0             // optional, for reproducible results
    }
}

⁠Image to Image Generation

{
    "input": {
        "prompt": "Description with image placeholders: <img><|image_1|></img>",
        "input_images": [
            "base64_encoded_image1",
            "base64_encoded_image2"  // optional, for multiple image inputs
        ],
        "height": 1024,              // optional, default: 1024
        "width": 1024,               // optional, default: 1024
        "guidance_scale": 3.0,       // optional, default: 3.0
        "img_guidance_scale": 1.6,   // optional, default: 1.6
        "separate_cfg_infer": false  // optional, set true if experiencing OOM
    }
}

⁠Response Format

{
    "output": {
        "image": "base64_encoded_output_image",
        "type": "text-to-image" // or "image-to-image"
    }
}

⁠Technical Details

• Base image: runpod/pytorch:2.1.1-py3.10-cuda11.8.0
• Key dependencies:
  • PyTorch >= 2.1.1
  • OmniGen (latest version)
  • CUDA 11.8 support
• Pre-downloaded model weights for instant startup

⁠Environment Requirements

• GPU with CUDA support recommended
• Minimum 24GB GPU memory
• If experiencing OOM (Out of Memory) errors:
  • Use separate_cfg_infer: true in the request
  • Consider reducing image dimensions
  • Adjust batch size if processing multiple requests

⁠About MockoFun

MockoFun⁠ is an advanced online photo editor and design tool. This image is part of the research for MockoFun's AI-driven image generation capabilities, enabling users to create and manipulate images with state-of-the-art AI technology.

⁠Tags

• latest: Most recent stable release
• x.x.x: Version-specific releases (when available)

⁠Support

For issues related to:

• MockoFun platform: Visit MockoFun Support⁠
• Image deployment: Create an issue in this repository
• OmniGen model: Visit OmniGen GitHub⁠

⁠License

This image is part of MockoFun's proprietary infrastructure. The underlying OmniGen model is subject to its original license terms.