AddNoise - ComfyUI Built-in Node Documentation

Complete documentation for the AddNoise node in ComfyUI. Learn its inputs, outputs, parameters and usage.

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

AddNoise

This node adds controlled noise to a latent image using specified noise parameters and sigma values. It processes the input through the model’s sampling system to apply noise scaling appropriate for the given sigma range.

How It Works

The node takes a latent image and applies noise to it based on the provided noise generator and sigma values. It first checks if there are any sigmas provided - if not, it returns the original latent image unchanged. The node then uses the model’s sampling system to process the latent image and apply scaled noise. The noise scaling is determined by the difference between the first and last sigma values when multiple sigmas are provided, or by the single sigma value when only one is available. Empty latent images (containing only zeros) are not shifted during processing. The final output is a new latent representation with the applied noise, with any NaN or infinite values converted to zeros for stability.

Inputs

Parameter	Data Type	Input Type	Default	Range	Description
model	MODEL	Required	-	-	The model containing sampling parameters and processing functions
noise	NOISE	Required	-	-	The noise generator that produces the base noise pattern
sigmas	SIGMAS	Required	-	-	Sigma values controlling the noise scaling intensity
latent_image	LATENT	Required	-	-	The input latent representation to which noise will be added

Outputs

Output Name	Data Type	Description
LATENT	LATENT	The modified latent representation with added noise