ComfyUI : ‘Use Everywhere’ Custom Nodes

Yushan777
5 min readDec 13, 2023

Simplify That Workflow!

I like ComfyUI but for me it’s a constant fight between keeping things neat and being able to quickly access a setting.

SDXL & Upscaler Workflow

Above is my basic SDXL base & Upscaler workflow. By using the Bus Node (a custom node by WASasquatch), it helps keep things organised. But as you can imagine, if i’m looking at the image previews, having to move back to the prompt nodes means scrolling back to the left and this can get old.

While I can hide the connections and arrange the nodes however I want, being able to see the control flow does help sometimes.

Use Everywhere (UE) nodes ‘broadcast’ data to any unconnected input that matches the data type and optional rules. Think of it in terms of using an ethernet cable vs Wifi.

To show you what I mean let’s take a look at a simpler workflow.

Above is a basic SD1.5 Workflow. It again uses the bus node. Let’s add a UE node that will accept 3 inputs, Anything Everywhere3.

you can also double-click the background and search for this node.

We will connect the MODEL, CLIP and VAE. The order does not matter since the input becomes anything that you connect to it.

But here is where the magic happens. Notice that when the CLIP is connected to the Everywhere node, the green and red CLIP inputs are highlighted? This means they are automatically receiving CLIP data broadcast from this Everywhere node. No more additional connections. This gives us room to rearrange nodes without having to worry about flow and order.

Now let’s add the Prompts Everywhere node. notice how the Bus node’s inputs have lit up? However, we can do away with the Bus node in this worklow since its job is really to act as a repeater. So let’s delete that.

So this is how the workflow looks now. Even though everything looks a bit disconnected, it works,

For the sake of this guide, let’s break it down a bit further. We will connect a node to the the Empty Latent Image node which will broadcast to the KSampler node. However notice how the VAE Decode node also takes Latent Image as an imput? We don’t want this to receive the latent from the empty latent space node, instead we want it to receive from the KSampler hence why it is directly connected to it. Finally we add one to the Image output from the the VAE Decode node which will broadcast to the Save Image node. Below is our finished Workflow.

If you wish to view the actual connections, you can click anywhere on the background and select Show UE links.

This will display an animated view of where the UE nodes are broadcasting to.

One final tweak… Let’s group the nodes.

Some say “forget the connections, arrange the nodes how you want or what makes sense to you”. Some say “I need to see the connections always!”

This is just another way. It removes a lot of connections and can result in more compact workflows that can be seen in a single view, reducing scrolling and panning around. But ultimately it is up to you.

Installing Use Everywhere Custom Nodes

The easiest way to install is via the ComfyUI Manager. However the caveat is that the manager itself is a third-party plugin too. If you are comfortable with the terminal or windows command prompt, go to your ComfyUI main directory:

# go to custom_nodes directory
cd custom_nodes

# clone the comfyui manager
git clone https://github.com/ltdrdata/ComfyUI-Manager.git

# Restart ComfyUI

start the ComfyUI server, at the bottom of the menu you should see the Manager button. Click it.

If you load up my workflow you can simply just click on Install Missing Custom Nodes, otherwise click on Install Custom Nodes. Search for Use Everywhere and you should find it. Install, Restart the ComfyUI Server.

Links:

Use Everywhere Custom Nodes

Resources from this guide

ComfyUI

If you have any questions or just want to learn more then join the
Stable Diffusion Training Discord Server. There are a lot smart people there who can help you.

--

--

Yushan777

No Click-Baity stuff! This is just my own journal so I can remember how I did stuff 🙂