3.7 KiB
title, description
| title | description |
|---|---|
| Code Interface | What are custom nodes and how are they used? |
Overview
A custom node is a combination of Python code and potentially some models weights. Custom nodes are extremely powerful, and allows the Comfy community to build their own functionality into ComfyUI.
If you prefer to read code, check out the example in the repository. Otherwise, we will explore one of Comfy's built-in nodes below.
Interface
Let's examine the interface of a custom node by looking at ComfyUI's built-in Load Checkpoint node. This loads a checkpoint file.
Functions
Every custom node can implement the following methods.
INPUT_TYPES
This defines the parameters that the custom node can take as an input.
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"ckpt_name": (folder_paths.
get_filename_list("checkpoints"),),
}}
Here you can see that input types is defined as a dictionary in Python. You can define inputs as required, hidden, or optional.
Each input must have a type (eg. VAE). If you use one of the special built-in types: INT, STRING, or FLOAT, then they can be lists.
default: You can define a default value.
min: You can define a min value.
max: You can define a max value.
step: You can use a slider.
IS_CHANGED
Optional. Allows the node to control when it is re-executed. ComfyUI attempts to only execute nodes that have changed in order to be more efficient.
Attributes
RETURN_TYPES
Tuple. The type of each element in the output tuple.
RETURN_TYPES = ("MODEL", "CLIP", "VAE")
RETURN_NAMES
Optional: The name of each output in the output tuple.
CATEGORY = "loaders"
FUNCTION
The name of the Python function in the custom node class to call.
For LoadCheckpointSimple, the function is defined like this:
FUNCTION = "load_checkpoint"
Entry Point Function
This Python function is called when the node is invoked. Needs to match the string under FUNCTION.
def load_checkpoint(self, ckpt_name, output_vae=True, output_clip=True):
ckpt_path = folder_paths.get_full_path("checkpoints", ckpt_name)
out = comfy.sd.load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, embedding_directory=folder_paths.get_folder_paths("embeddings"))
return out[:3]
OUTPUT_NODE
Bool. Defaults to False. Set to true if your node will output a result/image from the graph.
CATEGORY
String. The category you want the node to appear under the UI.
CATEGORY = "loaders"
WEB_DIRECTORY
Custom nodes can have custom UI.
This attribute sets the web directory. Any .js file in that directory will be loaded as a frontend extension.
Custom nodes can also include markdown documentation in the WEB_DIRECTORY/docs folder. See the Help Page section for details on how to add rich documentation for your nodes.
NODE_CLASS_MAPPINGS
A dictionary that contains all nodes you want to export along with their class names. Class names should be unique. This also means you can define multiple nodes in one "custom node" and export them all together.
NODE_CLASS_MAPPINGS = {
"CheckpointLoaderSimple": CheckpointLoaderSimple,
}
NODE_DISPLAY_NAME_MAPPINGS
If you want to define more human-readable names for each node.
NODE_DISPLAY_NAME_MAPPINGS = {
"CheckpointLoaderSimple": "Load Checkpoint",
}