Skip to main content
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
The GITSScheduler node generates noise schedule sigmas for the GITS (Generative Iterative Time Steps) sampling method. It calculates sigma values based on a coefficient parameter and number of steps, with an optional denoising factor that can reduce the total steps used. The node uses pre-defined noise levels and interpolation to create the final sigma schedule.

Inputs

ParameterData TypeRequiredRangeDescription
coeffFLOATYes0.80 - 1.50The coefficient value that controls the noise schedule curve (default: 1.20)
stepsINTYes2 - 1000The total number of sampling steps to generate sigmas for (default: 10)
denoiseFLOATYes0.0 - 1.0Denoising factor that reduces the number of steps used (default: 1.0)
Note: When denoise is set to 0.0, the node returns an empty tensor. When denoise is less than 1.0, the actual number of steps used is calculated as round(steps * denoise).

Outputs

Output NameData TypeDescription
sigmasSIGMASThe generated sigma values for the noise schedule