Workflows

Here are some ideas of how to get to know Comfy UI.

Example Workflows provided in ComfyUI

Go to Workflow > Browse Templates

Choose from a vide variety of templates.

Image Masking (Replace part of an Image)

If you want to change just a part of a picture, Comfy UI has a simple Workflow that lets you draw a mask.

1. Like above, go to Workflow > Browse Templates, then select Basics > Inpaint 
2. In the Load Image Node (see image above), click choose file to upload, then select a file from your local device.
3. Right-click on the Node, select Open in MaskEditor
4. Edit Mask, then click save
   

ControlNet

ControlNet allows for more precise control over img2img generation by conditioning the diffusion process on additional inputs like edge maps, depth maps, human poses, and more. Basically, it lets you "guide" the image generation more explicitly than with just a text prompt.

If you go to Workflow > Browse Templates, under the Tab ControlNet you will find a bunch of workflows to explore one functionality at a time.

The following tutorial with attached workflow to download lets you compare them in one file: https://openart.ai/workflows/oliviosarikas/lesson-9-controlnet---comfy-academy/eWSkFEeLjJ33zfCulBQs 

Download the workflow onto your local device, then go to Workflow > Open and select file.

Troubleshooting

When loading workflows, they might use nodes or models that you don't have installed. Here's a quick overview of how to deal with that.

Missing Node

Close the Missing Node Type notification, open Manager.

Go to Install Missing Custom Nodes 

1. Click install.
2. When installation is done, a pop up window will ask you to refresh the page. Click Refresh. 
3. You can only install one at a time. If you have multiple nodes to install, go back to Manager after refreshing and repeat the process.
4. Restart ComfyUI: Go back to Manager and press the red Restart Button.

Missing Models, missing Loras

Just close the Missing Loras notification. (The download Buttons will let you download the models onto your local device. Don't do that:) They're of little use there and you have already installed at least one model in the installation process)

In the node Load Checkpoint you can select your Stable Diffusion Model. Just click on the stepper button and select one of your installed models.

Same goes for missing Loras: In the node Load Lora, click on the stepper button lora_name and select a Lora. 

If you do want/need to add a Model or Lora, the easiest way is to do that in the Model Manager.
Other options are through the Jupiter Notebook in the installation process or by cloning a GitHub repository in Terminal.

Model Compatibility

In more complicated Workflows, make sure that the Lora and VAE are compatible with your stableStable Diffusion Model.

Model Versions are called v1-5, v2-1, SD_XL, ...