A nodes extension for use with InvokeAI.
NodeUtilities is a collection of utility nodes written by me over time.
To install, place the desired .py files into your InvokeAI invocations folder
located here:
Windows - invokeai\.venv\Lib\site-packages\invokeai\app\invocations\
Mac/Linux - invokeai/.venv/lib/python3.10/site-packages/invokeai/app/invocations/
Outputs an image with the background removed behind the subject using rembg.
Remove Background requires the rembg Python package.
To install rembg, first activate InvokeAI's virtual environment. From your
InvokeAI root folder execute: source .venv/bin/activate. Then execute:
pip install rembg.
Remove Background will pass through the unaltered original image if rembg
isn't installed in InvokeAI's virtual environment.
Note: rembg installs its onnx models to a hidden u2net directory in your user
home folder.
Outputs an image with some pixels made transparent. Uses a transparency threshold to identify pixels to be made transparent. Optionally crop to remaining pixels (subject) with a transparent border (px). Mask is form-fit to the subject in black by default. Rectangle mask draws a black bounding box around the subject to give painting more room to work with (when altering the subject). Invert the mask to alter the background. Rectangle mask is off when cropping to subject's bounding box (whole mask would be black). Cropped H&W are multiples of 8, and border is >=0 & a multiple of 8.
before:
after:
Adaptive Histogram Equalization using skimage.
Pad or crop an image's sides from the center by specified pixels. Positive values are outside of the image. Checkerboard added to background of result for demonstration. Also takes negative numbers for inward cropping.
Outputs a given value as a string, float, or integer. Rounds in the case of float to integer. Value input types include string, float, and integer.
Sleeps for a given interval in seconds. Optionally clears VRAM cache.
Overwrite a channel in the input image with a channel image (channel image is any image).
Blend/place a face into/onto an existing face. Downloads non-commercial
research models to invokeai/models/any/faceswapper. NOTE: requires
insightface library. Activate invokeai's venv. From the invokeai folder:
source .venv/bin/activate
Next, do:
pip install insightface
Face restoration using CodeFormer. Downloads non-commercial
research models to invokeai/models/any/face_restoration/.
Face restoration using GFPGAN. Downloads models to
invokeai/models/any/face_restoration/.
