If you feel overwhelmed by the nodes, you can hide the nodes and connections by using subgraphs. What does a subgraph do? ComfyUI subgraph: Game changer or a nice try?
Using subgraphs
Here’s the step-by-step guide on using subgraphs.
Step 1: Load a workflow
You can use any workflow with more than one node. We will use the following text-to-image < https://stable-diffusion-art.com/text-to-image/> workflow as an example. Download the workflow JSON file below.
Download < https://stable-diffusion-art.com/wp-content/uploads/2025/01/text-to-image-SD1.5-v3.json>
Drag and drop it to ComfyUI to load.
Step 2: Select some nodes
Select a group of nodes that you want to turn into a subgraph. We will select all nodes except the Save Image node. Hold down the Ctrl key and the left button of your mouse at the same time. Move the mouse to select all nodes except the Save Image node. You should see the nodes highlighted.
Step 3: Create a subgraph
Click the subgraph button on the floating toolbar.
You should see the selected nodes collapsed to a subgraph.
Drag the corner of the subgraph to enlarge it.Step 4: Select the visible settings
You see some settings, such as prompts and seed, visible on the subgraph. You can show or hide any settings.
Click the Edit Subgraph Widgets button on the floating menu.
A parameters menu appears. Click the eyes icon to show or hide any settings.
Rearrange the order by dragging a setting tag.
Edit the title of the subgraph by double-clicking the title.
You can download the final subgraphed workflow below.
Download < https://stable-diffusion-art.com/wp-content/uploads/2025/10/text-to-image-SD1.5-subgraph.json>
My take on ComfyUI subgraphs
ComfyUI subgraphs aim to address a major problem: the interface intimidates new users. Beginners often describe ComfyUI as having a steep learning curve
<https://www.reddit.com/r/comfyui/comments/1ld7006/struggling_with_comfyui_learning_curve_after_a111/> .
In reality, most people are content to run premade workflows (such as the workflow JSON files < https://stable-diffusion-art.com/workflows/> on this site).
Subgraph serves two purposes:
1. Hide nodes and lines to make the workflow look simpler.
2. Expose only necessary input fields.
Subgraphs make the local workflow functions more like those on Replicate <https://replicate.com/stability-ai/stable-diffusion> and fal <https://fal.ai/models/fal-ai/fast-sdxl> :
You don’t need to know how it works, and you can only change specific vital parameters. (e.g. prompt).
What needs to be improved?
I think subgraph is a step in the right direction. Here are a few things I wish would come true.
1. Widget labels: Take my Graffiti generator < https://stable-diffusion-art.com/graffiti-generator/> as an example. It uses a third-party Multiline Textbox node that didn’t display any label.
I want users to know that the first text box is for the graffiti text, the second is for the style, and so on. It would be great if I could assign a label to a widget, or if ComfyUI has a label node.
2. Widget placement. Currently, you can only place them linearly from top to bottom. More freedom would be great.
3. Restricting input range: The input ranges in the nodes themselves are often meant to be generic. Some values may not work for this workflow. For example, you don’t want to set the denoising strength < https://stable-diffusion-art.com/denoising-strength/> to 1 in some inpainting < https://stable-diffusion-art.com/inpainting/> workflows.
Okay, that’s all of my opinion, so, do you use subgraphs now or it will be later?
Good luck.











No comments:
Post a Comment
А что вы думаете по этому поводу?