OptimalStepsScheduler - ComfyUI Built-in Node Documentation

The OptimalStepsScheduler node creates a noise schedule (a sequence of sigma values) for use during diffusion sampling. It chooses the base noise levels from the selected model type, adjusts the schedule when denoising is only partially applied, and interpolates the levels so the returned sigmas match the requested step count.

Inputs

Parameter	Description	Data Type	Required	Range
`model_type`	The type of diffusion model to use for the noise level calculation. Each option uses its own predefined table of noise levels.	COMBO	Yes	”FLUX" "Wan" "Chroma”
`steps`	The total number of sampling steps to calculate (default: 20).	INT	Yes	3 to 1000
`denoise`	Controls the denoising strength, which adjusts the effective number of steps (default: 1.0).	FLOAT	Yes	0.0 to 1.0 (step: 0.01)

Note: The base noise level table for the selected model_type is resampled with log-linear interpolation whenever its length does not equal steps + 1, so the output always matches the requested step count. Note: When denoise is less than 1.0, the node uses round(steps * denoise) as the total number of effective steps and keeps only the matching tail of the schedule. If denoise is 0.0 or lower, the node returns an empty tensor.

Outputs

Output Name	Description	Data Type
`sigmas`	A sequence of sigma values representing the noise schedule for diffusion sampling. The final value in the sequence is always set to 0.	SIGMAS

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): fd48c94ca16c8a3d8e6f0138018e7b13c15d100d6147807bcb23d838899045b7

​Inputs

​Outputs

Inputs

Outputs