Eri - SD Forms: Beautiful AI Image & Video Generation

⚠️ SEMI FUNCTIONAL WORK IN DEVELOPMENT ⚠️

A beautiful, component-based visual workflow system for Stable Diffusion image and video generation with a stunning purple macOS-themed interface.

🎨 Features

• Beautiful Purple macOS Theme: Elegant LiteGraph.js interface with glass effects and smooth animations
• Component-Based Workflow: Drag-and-drop visual programming for AI generation
• Multiple Model Support: Flux, SD 1.5, SD 3.5, SDXL, Chroma, OmniGen
• Video Generation: AnimateDiff, LTX Video, WAN-VACE support
• Memory Optimized: Automatic CPU offload for large models
• Auto Model Detection: JSON-based configuration system

🚀 Quick Start

Prerequisites

• Python 3.8+
• CUDA-capable GPU (recommended)
• 16GB+ RAM

Installation

1. Clone the repository

git clone https://github.com/CodeAlexx/Eri.git
cd Eri

2. Install dependencies

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
pip install diffusers>=0.30.0 transformers>=4.38.0 accelerate
pip install fastapi uvicorn pillow numpy
pip install safetensors huggingface_hub

3. Configure model paths Edit model_paths_config.json to point to your model directories:

{
  "model_directories": {
    "flux": ["/path/to/your/models"],
    "sd15": ["/path/to/your/models"],
    "sdxl": ["/path/to/your/models"]
  }
}

4. Start the server

cd backend
python server.py

5. Open web interface Navigate to: http://localhost:8001

🎯 Generate Your First Image

1. Clear Canvas: Start with a fresh workspace

2. Add SDXL Model Component:

   • Drag an SDXL component from the right panel (currently the most stable option)

3. Generate Basic Workflow:

   • Click the Generate Basic Workflow button to automatically add and connect required components

4. Adjust Settings:

   • Modify steps (20-30 recommended)
   • Adjust CFG scale (7-8 works well)
   • Enter your prompt in the model component

5. Generate: Click the Generate button to create your first image

📁 Project Structure

Eri/
├── sd_forms/              # Core component system (THE GOLD!)
│   ├── components/        # Image/video generation components
│   ├── core/             # Pipeline system & registries
│   └── utils/            # Model path configuration
├── backend/              # FastAPI server
├── web/                  # LiteGraph.js frontend
├── model_paths_config.json # Model directory configuration
└── README.md

🔧 Key Components

Image Generation

• Flux Model: Fast 4-step generation
• SD 1.5: Classic Stable Diffusion
• SDXL: High-resolution 1024x1024
• Chroma: Flux-based model
• SD 3.5: Latest Stability AI model

Video Generation

• AnimateDiff: SD 1.5 animation
• LTX Video: Direct video generation
• WAN-VACE: Video generation model

Processing

• Sampler: Core generation component
• Upscaler: Image enhancement
• ControlNet: Guided generation
• VAE: Custom VAE loading

🎨 Beautiful Interface

The LiteGraph.js interface features:

• Purple gradient theme with glass morphism
• Smooth animations and hover effects
• macOS-style components and typography
• Draggable property panels
• Auto-connecting workflow nodes

⚙️ Configuration

Memory Optimization

Large models automatically use CPU offload. Disable in component properties if you have enough VRAM.

Model Auto-Detection

The system automatically finds models in:

• SwarmUI model directories
• Custom paths in model_paths_config.json
• HuggingFace cache

🐛 Known Issues

• This is development software - expect bugs!
• Large models require significant RAM/VRAM
• Some components may need additional testing
• Error handling is still being improved

🤝 Contributing

This is an active development project. Feel free to:

• Report issues
• Suggest improvements
• Submit pull requests
• Share your workflows

📄 License

Open source - use at your own risk during development phase.

---

Generate beautiful images and videos with the power of visual programming!

Built with love for the AI art community 💜