What is ComfyUI?

ComfyUI serves as a node-based graphical user interface for Stable Diffusion. Users have the ability to assemble a workflow for image generation by linking various blocks, referred to as nodes. These nodes include common operations such as loading a model, inputting prompts, defining samplers and more. ComfyUI dissects a workflow into adjustable components, enabling users to customize their own unique processes.

A ComfyUI default workflow on ThinkDiffusion with the ThinkDiffusionXL model loaded. The output image shows a galaxy in a bottle.

Nodes and why it's easy

Nodes work by linking together simple operations to complete a larger complex task. It might seem daunting at first, but you actually don't need to fully learn how these are connected. The way ComfyUI is built up, every image or video saves the workflow in the metadata, which means that once an image has been generated with ComfyUI, you can simply drag and drop it to get that complete workflow. No need to connect anything yourself if you don't want to!

In this post we'll show you some example workflows you can import and get started straight away.

A ComfyUI workflow in ThinkDiffusion showcasing how an image of a galaxy in a bottle is made

Easy starting workflow

These versatile workflow templates have been designed to cater to a diverse range of projects, making them compatible with any SD1.5 checkpoint model. The initial set includes three templates:

  1. Simple Template
  2. Intermediate Template
  3. Advanced Template

Primarily targeted at new ComfyUI users, these templates are ideal for their needs.

SD1.5 Template Workflows for ComfyUI - v2.0 | Stable Diffusion Workflows | Civitai
*** Update 21/08/2023 - v2.04 Fixed missing Seed issue plus minor improvements *** These workflow templates are intended as multi-purpose templates…

Searge's Advanced SDXL workflow

SeargeXL is a very advanced workflow that runs on SDXL models and can run many of the most popular extension nodes like ControlNet, Inpainting, Loras, FreeU and much more. It's running custom image improvements created by Searge and if you're an advanced user, this will get you a starting workflow where you can achieve almost anything when it comes to still image generation. This workflow is not for the faint of heart, if you're new to ComfyUI, we recommend selecting one of the simpler workflows above.

You will only need to load the .json file or drag & drop an image. All the custom nodes come pre-installed with ThinkDiffusion.

Searge's advanced ComfyUI workflow loaded in ThinkDiffusion
Searge-SDXL: EVOLVED v4.3.1 - Optimized Workflow for ComfyUI - 2023-10-26 - txt2img, img2img, inpaint, revision, controlnet, loras, FreeU v1 & v2, ... - v4.3.1 | Stable Diffusion Workflows | Civitai
Searge-SDXL: EVOLVED v4.3.1 Version 4.x is here. I made a convenient install script that can install the extension and workflow, the python depende…

Make your own animations with AnimateDiff

AnimateDiff is a powerful tool to make animations with generative AI. In these ComfyUI workflows you will be able to create animations from just text prompts but also from a video input where you can set your preferred animation for any frame that you want.

An animation of a muscular man with white hair made with AnimateDiff in ThinkDiffusion
An animation of a woman with long hair under cherry blossom trees made with AnimateDiff in ThinkDiffusion
[GUIDE] ComfyUI AnimateDiff Guide/Workflows Including Prompt Scheduling - An Inner-Reflections Guide | Civitai
Introduction AnimateDiff in ComfyUI is an amazing way to generate AI Videos. In this Guide I will try to help you with starting out using this and…

How to use AnimateDiff

Load the workflow, in this example we're using Basic Text2Vid

Set your number of frames. Depending on your frame-rate, this will affect the length of your video in seconds. It will always be this frame amount, but frames can run at different speeds. For example, 50 frames at 12 frames per second will run longer than 50 frames at 24 frames per second.

a ComfyUI node in ThinkDiffusion that selects the amount of frames for AnimateDiff

Set your desired size, we recommend starting with 512x512 or 768x768.

a ComfyUI node in ThinkDiffusion that changes the width and height of animation output from AnimateDiff

Select your desired model, make sure it's an 1.5 model. That means no model named SDXL or XL. Dreamshaper is a good starting model.

a ComfyUI node in ThinkDiffusion that selects the model loaded for AnimateDiff

Set your desired positive and negative prompt (this is what you want, and don't want, to see)

a ComfyUI node in ThinkDiffusion where you can write the prompt for AnimateDiff

Set your desired frame rate and format (gif, mp4, webm). Pingpong means that in the end, the animation will reverse and keep looping by reversing.

a ComfyUI node in ThinkDiffusion where you combine output frames into an AnimateDiff video

Start your generation by pressing Queue Prompt!

A button saying queue prompt from ComfyUI in ThinkDiffusion

Get creating!

ComfyUI and these workflows can be run on a local version of SD but if you’re having issues with installation or slow hardware, you can start creating today in your browser with ThinkDiffusion.

Oh, and share your art!

So go out there and create! And while you're at it, come join our Discord server and share your art for everyone to enjoy https://discord.gg/Wg2VExCrnS