ComfyUI offers a flexible alternative to some of the more common UIs for AI generation, such as AUTOMATIC1111. With ComfyUI, you can easily set up and run various models, including the highly popular Flux model. Below, we'll walk you through how to set up and run different variants of Flux on ComfyUI, with steps for downloading the model files, updating ComfyUI, and running workflows.

Running models locally allows you to keep all data processing within your own system, ensuring privacy and avoiding reliance on external servers. Additionally, local setups reduce ongoing cloud costs and can offer a more customizable experience for AI enthusiasts or professionals working with complex workflows.

The Flux model comes in several different variants, depending on your hardware capabilities and quality needs:

Start by downloading the Flux1 dev FP8 model checkpoint:

Once downloaded, place the model file in the following folder: ComfyUI > models > checkpoints.

If you haven't already updated ComfyUI, now is the time to do so. ComfyUI introduced native support for Flux in August 2024, so make sure you are running the latest version. Use the ComfyUI Manager for a quick update:

Go to Manager > Update All and reload the page to apply the changes.

Now download the corresponding workflow file:

Drop the JSON workflow file into ComfyUI. If you encounter any red error boxes, reload the interface to resolve the issue.

You're all set! Just press Queue Prompt to generate images using the Flux FP8 model.

For faster sampling at slightly reduced quality, you can opt for the Flux Schnell version. Download the Schnell model here:

Place this model file in ComfyUI > models > unet.