ComfyUI UNET Loader: Load Diffusion Model Node Guide

ComfyUI Wiki

UNET Loader in ComfyUI (now called Load Diffusion Model) loads a model by unet_name from models/unet, with weight_dtype options default, fp8_e4m3fn, fp8_e4m3fn_fast and fp8_e5m2.

comfyUI节点-UNETLoader|UNET加载器

This node has been renamed as Load Diffusion Model.

comfyUI节点-Load-Diffusion-Model|UNET加载器

Documentation

  • Class name: UNETLoader
  • Category: advanced/loaders
  • Output node: False

The UNETLoader node is designed for loading U-Net models by name, facilitating the use of pre-trained U-Net architectures within the system.

Input types - UNET Loader Guide | Load Diffusion Model

ParameterComfy dtypeDescription
unet_nameCOMBO[STRING]Specifies the name of the U-Net model to be loaded. This name is used to locate the model within a predefined directory structure, enabling the dynamic loading of different U-Net models.
weight_dtype...🚧 fp8_e4m3fn fp9_e5m2

Output types - UNET Loader Guide | Load Diffusion Model

ParameterComfy dtypeDescription
modelMODELReturns the loaded U-Net model, allowing it to be utilized for further processing or inference within the system.

Load Diffusion Model Workflow Example | UNET Loader Guide

  1. Install the UNET models
  2. Download the workflow file
  3. Import workflow in comfyUI
  4. Choose the UNET model and run the workflow

Download FLux.1 UNET Model

FLUX.1-schnell on hugging face

File NameSizeLinkNote
flux1-schnell.safetensors23.8GBDownloadFor lower memory usage
flux1-dev.safetensors23.8GBDownloadIf you have high VRAM and RAM.
  1. Downloaded the flux1-schnell.safetensors
  2. Place downloaded model files in ComfyUI/models/unet/ folder

Where you can download UNet models and how to install it?

  1. You can download UNet models from this list: Unet Model Resource
  2. After download the model files, you should place it in /ComfyUI/models, then refresh ComfyUI or restart it. If everything is fine, you can see the model name in the dropdown list of the UNETLoader node.

Comments

Sign in with GitHub to join the discussion.

Loading comments…