KSampler

Documentation

• Class name: KSampler

• Category: sampling

• Output node: False

The KSampler node is designed for advanced sampling operations within generative models, allowing for the customization of sampling processes through various parameters. It facilitates the generation of new data samples by manipulating latent space representations, leveraging conditioning, and adjusting noise levels.

Input types

Parameter	Comfy dtype	Description
model	MODEL	Specifies the generative model to be used for sampling, playing a crucial role in determining the characteristics of the generated samples.
seed	INT	Controls the randomness of the sampling process, ensuring reproducibility of results when set to a specific value.
steps	INT	Determines the number of steps to be taken in the sampling process, affecting the detail and quality of the generated samples.
cfg	FLOAT	Adjusts the conditioning factor, influencing the direction and strength of the conditioning applied during sampling.
sampler_name	COMBO[STRING]	Selects the specific sampling algorithm to be used, impacting the behavior and outcome of the sampling process.
scheduler	COMBO[STRING]	Chooses the scheduling algorithm for controlling the sampling process, affecting the progression and dynamics of sampling.
positive	CONDITIONING	Defines positive conditioning to guide the sampling towards desired attributes or features.
negative	CONDITIONING	Specifies negative conditioning to steer the sampling away from certain attributes or features.
latent_image	LATENT	Provides a latent space representation to be used as a starting point or reference for the sampling process.
denoise	FLOAT	Controls the level of denoising applied to the samples, affecting the clarity and sharpness of the generated images.

Output types

Parameter	Comfy dtype	Description
latent	LATENT	Represents the latent space output of the sampling process, encapsulating the generated samples.