antrobots ComfyUI Nodepack

welcome to my node pack! This pack serves as a compilation of all the custom nodes I have and will create, with the goal of implementing functionality that I find lacking or obtuse. If you experience any issues with this node pack, please make a github issue! any help finding potential bugs or errors is greatly appreciated.

Below is a description of what each node in this node pack does.

Sampling Nodes

KSampler (advanced) with denoise

this node is a variant of the built-in KSampler (advanced) with the additional ability to provide a denoise value. This allows you to swap models mid generation in img-to-img workflows much more efficiently. denoise works in tandem with start step and end step.

Ksampler with Refiner

This node is an extension of the Ksampler (advanced) node that takes in two sets of model parameters. It will then generate with base_model until refine_step, in which it will switch to refine_model (re-encoding the latent with refine_vae if the two passed in VAEs are different). it also has an optional mask can also be provided, which will set the latent noise mask if given.

Flow Control nodes

Swap

This node takes two inputs of any type. If the swap bool is False, it simply pipes input 1 into output A and input 2 into output B. however, if the swap bool is True, it swaps the outputs, mapping 1 to B and 2 to A. This is useful when combined with piping workflows, as it allows you to easily and quickly switch to a different input pipe or swap two pipes around with a boolean.

Image Handling Nodes

Crop Image and Mask:

This node is used to crop an image and its corresponding mask. It takes an image, a mask, a minimum width, a minimum height, and padding as inputs. It scales the mask to the image, converts the mask into a region, scales the box to the image, and crops the image and mask. If the mask is empty, it returns the input image and mask as they are.

Scale Image to Size:

This node is used to scale an image to a desired size. It takes an image, a desired size, and a boolean value doLarger as inputs. It calculates the scale factor based on the desired size and the smaller or larger dimension of the image (depending on the doLarger value), scales the image to the final size, and returns the scaled image.

Paste with Masks:

This node is used to paste a source image onto a destination image using masks. It takes a source image, a source mask, a destination image, and a destination mask as inputs, as well as a horizontal padding, a vertical padding, and a global padding. It treats the source mask as an Alpha channel for the source image. it then pastes the source image onto the destination image using the destination mask as the area to paste onto, cropping and resizing if necessary. The output mask and image are padded with the horizontal and vertical padding values. Global padding can also be used to pad both width and height at once.

Alpha Composite:

This node is used to composite a source image and a destination image using alpha blending. It takes a source image, a destination image, a source mask, a destination mask, an x offset, and a y offset as inputs. It pastes the source image onto the destination image with offsets, treating the respective masks as alpha channels. Both source mask and destination mask are optional. If either is not provided, the respective image will be pasted as is.

##Dict Nodes These are various nodes that can pack information into dicts to be sent around and parsed as an alternative workflow to piping.

Set Dict

This is the core of how these nodes work. any value put into this node will be packaged into a DICT labeled as key and sent out. A dict can optionally be passed in, in which case it will instead append the value to the input dict.

Get Dict

This node takes in a DICT, a key, and a default value as inputs. if key is found in the DICT, it will pass it out to the * output. otherwise, default will be used

Get Dict TYPE

This entry describes all of the nodes that follow this format, as their core functionality is the same. These nodes behave similar to thee standard Get Dict, with two exceptions.

1. default is optional. if default is not provided, a standard default for the type will be provided automatically.
2. inputs are validated with their type. If the input is not of the correct type, default will be used instead.

Get Dict Unsafe

This is a special variant of get dict with NO default. If the input does not exist, it will return None. This can be useful if you are using a lot of custom types and don't want to populate all the defaults. Note that if the key does not exist, most workflows will likely break, so use with caution.