SamplerLCMUpscale - ComfyUI Built-in Node Documentation

The SamplerLCMUpscale node provides a specialized sampling method that combines Latent Consistency Model (LCM) sampling with image upscaling capabilities. It upscales the image progressively during the sampling process using various interpolation methods, allowing higher-resolution outputs to be generated in a single sampling pass. The output is a configured sampler object that can be connected to a sampling node.

Inputs

Parameter	Description	Data Type	Required	Range
`scale_ratio`	The total scaling factor to apply during upscaling. A value of 1.0 keeps the original resolution (default: 1.0)	FLOAT	Yes	0.1 - 20.0
`scale_steps`	The number of steps to use for the upscaling process. Use -1 for automatic calculation based on the sampling schedule (default: -1)	INT	Yes	-1 - 1000
`upscale_method`	The interpolation method used for upscaling the image at each upscaling step (default: “bislerp”)	COMBO	Yes	”bislerp" "nearest-exact" "bilinear" "area" "bicubic”

scale_ratio and scale_steps are advanced parameters. The image is upscaled gradually from its original size to the target scale_ratio over the upscaling steps. When scale_steps is -1, the number of upscaling steps is calculated automatically as about half the number of sampling steps, with a minimum of 2; when a positive value is given, the node adjusts it internally and limits it based on the total number of sampling steps.

Outputs

Output Name	Description	Data Type
`sampler`	A configured sampler object that performs LCM sampling with progressive upscaling, ready to be used in the sampling pipeline	SAMPLER

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

Source fingerprint (SHA-256): 5d6f6472fbb4d2c66a8a8b9d6dc34dcc52ac8272589fd6c29e4084d6cab3141b

​Inputs

​Outputs

Inputs

Outputs