Key features of ComfyUI include:
Support for Stable Diffusion 1.x and 2.x
Supports various model combinations, such as Upscaler, Denooiser, etc.
Provides workflow saving and loading functions
Generate a complete workflow from PNG in one click
Runtime optimization, re-execute only the modified parts
The principle of ComfyUI
ComfyUI is based on a graphical node editor. By dragging various nodes onto the canvas and connecting the relationships between the nodes, you can build an end-to-end Workflow.
Each node represents a Stable Diffusion-related model south korea mobile database or function, and the nodes transmit image information through connections, ultimately forming a workflow from image generation to processing.
For example, a simple upsampling workflow might include:
Text prompt node - input text prompt
SD node - Stable Diffusion generated images
ESRGAN node - Upsample images using ESRGAN
When the Workflow is built, ComfyUI will call the APIs of each model in the background, execute each node in the order of the process, and achieve automatic result output.