One of ComfyUI’s greatest strengths as a diffusion model platform is its rich custom node ecosystem. Many talented developers have written their own custom nodes that greatly expand the capabilities of ComfyUI, from image processing and animation to general usability and ergonomics.
Because there are so many custom nodes of varying quality and support it can feel intimidating to get started working with them. To understand this ecosystem better, we asked our good friends at Comfy Deploy to pull the top 5 most used custom node packs across the thousands of deployed ComfyUI workflows on their platform.
If you’re someone who’s always wanted to try out ComfyUI custom nodes but don’t know where to start, this list is for you.
| Pack | Description | |
|---|---|---|
| 1 | WAS Node Suite | Hundreds of nodes across image processing and more |
| 2 | ComfyUI Impact Pack | Image enhancement nodes especially for face detailing |
| 3 | ComfyUI IPAdapter Plus | Style transfer of reference images |
| 4 | ComfyUI Essentials | Quality-of-life improvement nodes |
| 5 | KJNodes for ComfyUI | Quality-of-life and simple image transformation nodes |
In this post, we’ll go through each of these packs and provide a simple:
- Description and use case
- Modal App that builds the image with the custom node pack’s dependencies and serves an interactive ComfyUI session
- JSON workflow you can drag and drop into the session
To run each example:
- Run
modal serve <custom_node>_example.py - Click the UI link after the image builds to open an interactive ComfyUI web session
- Drag and drop the JSON workflow into the UI
#0. ComfyUI Manager
It goes without saying you need ComfyUI Manager if you want to run custom nodes. ComfyUI Manager is the primary entrypoint for most ComfyUI users for discovering and installing custom nodes, so it deserves the first spot on any list.
Keep in mind that if you’re using ComfyUI with Modal, the ComfyUI Manager Menu does not work for downloading custom nodes because your filesystem is not persisted across sessions. That being said, our ComfyUI example uses comfy-cli, which uses ComfyUI Manager CLI in the backend to manage the custom node installation.
#1. WAS Node Suite
WAS Node Suite is the most popular ComfyUI custom node pack and contains hundreds of nodes across image processing, prompt processing, and general workflow improvements.
In the example below, we use WAS Node Suite’s Image Generate Gradient, Image Gradient Map, and Image Blending Mode nodes to apply a warm, autumn-y filter (adapted from this example).
Sample code here.
#2. ComfyUI Impact Pack
ComfyUI Impact Pack from ltdrdata (also the creator of ComfyUI Manager) is another popular extension focused on image enhancement, especially around face detailing and image segmentation.
In the example below (adapted from this workflow), we first generate an AI image with the prompt “A smiling woman with short hair”.
In classic AI art fashion, we get a (slightly horrifying) distorted face. Now let’s pass the image through the FaceDetailer node to make it look more realistic.
Much better. This example also shows how FaceDetailer can crop the headshot and segment out the face when the UltralyticsDetectionProvider and SAMLoader nodes are passed in.
Sample code here.
#3. ComfyUI IPAdapater Plus
ComfyUI IPAdapter Plus is the ComfyUI implementation of IP-Adapter (Image Prompt adapter) which allows you to pass reference images to the image generation process to achieve style transfer e.g. “x in the art style of y”.
In the example below (adapted from this video), we use the IPAdapter Unified Loader and IPAdapter nodes with the weight_type set to “style transfer” to generate a painting of a golden retriever in the style of “Starry Night”:
Sample code here.
#4. ComfyUI Essentials
ComfyUI Essentials (developed by the same author as ComfyUI IPAdapter Plus) contains mostly quality-of-life workflow improvement nodes:
- Printing image dimensions to the console (Get Image Size, Console Debug)
- Simple image transformations (Image Resize, Image Crop, Image Flip)
- Simple mask transformations (Mask Blur, Mask Flip)
The workflow below (adapted from this workflow) gives a quick glance into all of the Essentials nodes:
Sample code here.
#5. KJNodes for ComfyUI
KJNodes for ComfyUI contains quality-of-life and simple image transformation nodes. For instance, the Color Match node allows you to apply a reference image’s color palette to a target image.
In the example below, I applied the iconic red hallway palette from “In the Mood for Love” to a picture I took of a Portuguese tram.
Sample code here.
Conclusion
The rapid development of the ComfyUI custom node ecosystem historically has made the developer experience painful, but thankfully there have been a lot of recent improvements:
- Comfy Org’s work establishing the Comfy Registry to standardize custom node development
- Tools like Comfy Deploy that make it easy to productionize a ComfyUI workflow
- Serverless compute platforms like Modal that allow you to programmatically define your ComfyUI environment (models, custom nodes) and iterate quickly while only paying for the compute you use
Make you sure to check out our official ComfyUI example which you can combine with these simple code snippets to serve your workflow via API.
