Extensions/ComfyUI-Universal-Latent
ComfyUI Extension

ComfyUI-Universal-Latent

Enhanced empty latent node with extended aspect ratio support for ComfyUI

By theshubzworld·Created 8 months ago·Updated 8 months ago· 1
theshubzworld/ComfyUI-Universal-Latent
Nodes1
On cloudLocal install
Categorylatent
Stars1
Updated8 months ago
Readme

🌌 Universal Latent Node for ComfyUI

An enhanced empty latent node for ComfyUI with extended aspect ratio support and optimized performance. Compatible with various models including SD3.5 and SDXL.

✨ Features

  • Wide Range of Aspect Ratios - Supports square, portrait, landscape, and custom aspect ratios
  • High Resolution Support - Up to 1536x1536 resolution
  • Aspect Ratio Lock - Lock width or height while adjusting the other dimension
  • Flexible Downsampling - Auto or manual selection of downsampling factor
  • Batch Processing - Generate multiple latent tensors at once
  • Optimized Performance - Efficient tensor operations for faster processing

🚀 Installation

  1. Clone this repository into your ComfyUI custom_nodes directory:

    cd ComfyUI/custom_nodes
    git clone https://github.com/theshubzworld/ComfyUI-Universal-Latent.git
    
  2. Restart ComfyUI

🎨 Usage

  1. Find the node in the node browser as "🌌✨ Universal Latent (Enhanced) 🎨⚡"
  2. Select your desired resolution from the dropdown
  3. Adjust batch size and other parameters as needed
  4. Connect the output to your model's latent input

⚙️ Parameters

  • Resolution: Predefined resolutions with common aspect ratios
  • Batch Size: Number of latent tensors to generate
  • Width/Height Override: Manually set dimensions
  • Aspect Ratio Lock: Lock width or height while adjusting the other dimension
  • Latent Channels: Number of channels in latent space (4 for SD, 16 for Flux)
  • Downsample Factor: Auto (8) or manual selection (4 or 8)
  • Invert Ratios: Swap width and height dimensions

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.