Flip Sigmas

Documentation

  • Class name: FlipSigmas

  • Category: sampling/custom_sampling/sigmas

  • Output node: False

The FlipSigmas node is designed to manipulate the sequence of sigma values used in diffusion models by reversing their order and ensuring the first value is non-zero if originally zero. This operation is crucial for adapting the noise levels in reverse order, facilitating the generation process in models that operate by gradually reducing noise from data.

Input types

Parameter
Comfy dtype
Description

sigmas

SIGMAS

The ‘sigmas’ parameter represents the sequence of sigma values to be flipped. This sequence is crucial for controlling the noise levels applied during the diffusion process, and flipping it is essential for the reverse generation process.

Output types

Parameter
Comfy dtype
Description

sigmas

SIGMAS

The output is the modified sequence of sigma values, flipped and adjusted to ensure the first value is non-zero if originally zero, ready for use in subsequent diffusion model operations.

Last updated