Enhance your ComfyUI experience with this collection of experimental nodes.
While the selection is currently limited, expect more additions in the future. Please note that these nodes are in an experimental stage and may contain bugs or issues.
A comprehensive Checkpoint VAE loader with additional features such as a clip skip value selector and initial empty latent image generation.
Combined positive and negative prompts, designed to save space.
Utilizes PIL for simple image manipulation. Currently, batch image processing is not supported, but development is ongoing.
Replaces a keyword from input text with a random line from a text file. If no input is given, it will output just a random line. It can now return multiple random lines from the same file, but be cautious as duplicates may occur. Wildcards files are stored under the extras/chibi-wildcards directory.
Appends the embedding text to the end of the input text ", (embedding:filename.pt:weight) ".
Text in, Conditioning out.
A 4-in-4-out node for text conditioning.
This feature merges the functionalities of VaeDecode and SaveImage, offering three distinct modes for filenames:
- Timestamp: Saves files to the standard output directory with a Unix timestamp as the filename (e.g., 1698462650001.png).
- Fixed: Utilizes the
fixed_filename
variable for individual images and batches (e.g., output_001.png). - Fixed Single: Uses the
fixed_filename
variable but saves all images under the same filename, not ideal for batches (e.g., output.png).
A simple text box. If pass-through text is supplied, it updates the textbox contents and sends it forward.
Displays the resolution of the input image and sends it along with two int values for the dimensions.
A quick way to resize a large image to a size your GPU can handle:
- "largest" sets the biggest dimension to the specified size while maintaining aspect ratio.
- "smallest" sets the lowest dimension to the specified size while maintaining aspect ratio.
- "all" sets all dimensions to the specified size.
- "height" and "width" only set the chosen dimension.
Converts a number into a text string, primarily for debugging purposes.
Adds text to a image or generates a blank image with text, also outputs a mask for the text, Ubuntu fonts have been included under extras.
Ksampler but less things to change because it uses presets, what could go wrong?
Could possibly be a bad idea, and still being worked on.
Added some functionality that my ai-info project has to get metadata from images generated by comfyui or automatic1111 (could be buggy with these), Also can make latents if supplied with a VAE, and the filename,width,height can be output.
Basic random or fixed number generator
Splits text at the first separator found from the front or back (reverse) of the text, you can choose which half you want output.
To install, download the repository to the custom_nodes directory:
cd custom_nodes
git clone https://github.com/chibiace/ComfyUI-Chibi-Nodes/
If you want to see latent image previews in ComfyUI's Sampler nodes append the following when launching the server
--preview-method auto
eg. python main.py --preview-method auto