ComfyUI Tensor Operations Nodes

This repo contains nodes for ComfyUI that implement some helpful operations on tensors, such as normalization.

Updates

February 3rd, 2025

• Updated Fast Image to Noise to provide an optional attention_mask input, which allows you to specify which pixels you wish to sample for the color information, ignoring unmasked pixels. See the new documentation below for more details.

June 12th, 2024

• Added Fast Image to Noise node, which generates a new image that is effectively a remix of the pixels in a source image.
• This is all done on the GPU and is lightning fast in comparison with the WAS Node Suite Image to Noise node.

February 2nd, 2024

• Initial release. Added Image Match Normalize and Latent Match Normalize nodes.

Nodes

Image Match Normalize

This node returns a normalized version of the target image using the mean and standard deviation of each color channel of the source image. If you want the color and brightness of your image to match the colour and brightness of another image, this is the node for you.

Latent Match Normalize

This node returns a normalized version of the target latent using the mean and standard deviation of each channel of the source latent. Latents encode color information differently than images and you may find that normalizing an image by instead normalizing its latent representation results in a "better" result that is closer to the coloring of the source.

Fast Image to Noise

Generates a new image by remixing the pixels of the source image randomly, with several customization options:

• num_colors: Number of colors to sample for the noise palette (2-256)
• black_mix: Probability of replacing pixels with black (0-1)
• gaussian_mix: Amount of Gaussian blur to apply (0-1024)
• brightness: Brightness adjustment factor (0-2)
• mask: Optional mask input that restricts color palette selection to masked areas only
• output_mode: Choose between "batch" or "list" output format
• seed: Random seed for reproducible results

The node is intended to match the functionality of the Image to Noise node in the WAS Node Suite but operates entirely on the GPU for significantly faster processing. When a mask is provided, the node will only sample colors from the masked (white) areas of the image to create the noise palette, allowing for more targeted noise generation.

This node is useful for inpainting applications where you need to match the color of some part of an image. Use a node such as ComfyUI Segment Anything to automatically select the part of the image you wish to sample color from and feed its outputs into Fast Image to Noise (both the image and the mask). Then you can apply the color-matched noise output onto the area of the image you wish to inpaint using another mask. For example, you might use ComfyUI Segment Anything to select the color palette from the "sofa" in an image of a living room and then paint that over the "chair" before running your inpainting sampling to finish off the inpainting of the chair.