This repository provides custom nodes and integrated workflows for Stable Diffusion XL (SDXL) within the ComfyUI ecosystem. It aims to simplify and enhance the SDXL generation process for users by offering a unified workflow that supports text-to-image, image-to-image, and inpainting modes, along with advanced features like multi-LoRA support and ControlNet integration.

How It Works

The project offers a comprehensive, single-file JSON workflow designed for ComfyUI, which leverages custom nodes for advanced functionality. It features a highly optimized processing pipeline, reportedly up to 20% faster than previous versions. Key architectural choices include integrated support for up to five LoRAs simultaneously, multiple ControlNet models, and improved high-resolution upscaling methods that replace the traditional "Hi-Res Fix." The workflow is designed with a consolidated interface for easier parameter adjustment.

Quick Start & Requirements

• Installation: Recommended for Windows users via a provided installer script (SeargeSDXL-Installer.bat) placed in the ComfyUI portable directory. Manual installation involves git clone into ComfyUI/custom_nodes/ and running python -m pip install opencv-python.
• Prerequisites: ComfyUI (portable version recommended for Windows installer), Python.
• Models: Requires downloading and placing specific SDXL base and refiner checkpoints, VAE, upscaler models, ControlNet models, and annotator models into the appropriate ComfyUI model directories. Links to Huggingface are provided.
• Resources: SDXL base model is 7GB, refiner is 6GB. ControlNet LoRAs are substantial (774MB each).

Highlighted Details

• Supports up to 5 LoRAs and 5 ControlNet models simultaneously.
• Features optimized processing pipeline, up to 20% faster.
• Includes improved high-resolution modes and FreeU v1/v2 support.
• Generates metadata embeddings in images for workflow reproducibility.

Maintenance & Community

The project is actively maintained, with recent updates (v4.3.2) focusing on ComfyUI compatibility. Changelogs detail new features and bug fixes across versions. Links to community resources are not explicitly provided in the README.

Licensing & Compatibility

The repository is licensed under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

Some features from previous versions (v3.4) like Prompt Styling, advanced Prompting Modes, and Condition Mixing are not yet included in the v4.x architecture but are planned for future releases. Users must ensure they are using the latest workflow JSON file with the corresponding custom node version.