SpectrumSEAKSamplerModGuidance

Enhance sampling with modulation guidance and spectral techniques for quality-driven results in SpectrumKSampler family.

SpectrumSEAKSamplerModGuidance:

The SpectrumSEAKSamplerModGuidance node is a sophisticated tool designed to enhance the sampling process by integrating modulation guidance with spectral evolution-aware techniques. This node is part of the SpectrumKSampler family, which is known for its ability to optimize the sampling process by leveraging advanced spectral methods. The primary goal of this node is to provide a more efficient and quality-driven sampling experience by incorporating modulation guidance that steers the sampling process towards desired quality tags. This is achieved through a combination of adaptive spectral techniques and modulation parameters that allow for fine-tuning of the sampling process. The node is particularly beneficial for AI artists looking to achieve high-quality outputs with improved detail and efficiency, as it combines the strengths of modulation guidance with spectral forecasting to deliver superior results.

SpectrumSEAKSamplerModGuidance Input Parameters:

model

The model parameter refers to the AI model that will be used for the sampling process. It is crucial as it determines the underlying architecture and capabilities that the node will leverage during sampling. There are no specific minimum or maximum values, as this parameter is typically a pre-trained model object.

seed

The seed parameter is used to initialize the random number generator, ensuring reproducibility of results. By setting a specific seed value, you can achieve consistent outputs across different runs. The default value is often set to a random number, but you can specify any integer to control the randomness.

steps

This parameter defines the number of steps the sampling process will take. More steps generally lead to higher quality outputs but at the cost of increased computation time. The minimum value is typically 1, with no strict maximum, though practical limits are imposed by computational resources.

cfg

The cfg parameter stands for configuration and is used to adjust the strength of the guidance during sampling. It influences how closely the sampling process follows the desired quality tags. The default value is usually set to a moderate level, allowing for balanced guidance.

sampler_name

This parameter specifies the name of the sampler to be used. It determines the algorithmic approach for sampling, with options varying based on the implementation. There are no specific minimum or maximum values, as it is a string identifier.

scheduler

The scheduler parameter controls the scheduling strategy for the sampling process. It affects how the steps are distributed over time, impacting the efficiency and quality of the output. The default value is often set to a standard scheduler, but advanced users can specify custom strategies.

positive

The positive parameter contains the positive prompts or conditions that guide the sampling process towards desired outcomes. It is a crucial input for steering the model in the right direction. There are no specific minimum or maximum values, as it is typically a text or feature vector.

negative

This parameter contains the negative prompts or conditions that the sampling process should avoid. It helps in refining the output by steering away from undesired features. Similar to the positive parameter, it is usually a text or feature vector without strict value limits.

latent_image

The latent_image parameter represents the initial latent space representation of the image to be sampled. It serves as the starting point for the sampling process. There are no specific minimum or maximum values, as it is typically a tensor or array.

adapter

The adapter parameter is used to specify any additional model adapters that modify the behavior of the sampling process. It allows for customization and fine-tuning of the model's response. There are no specific minimum or maximum values, as it is typically a model component.

quality_tags

This parameter contains the quality tags that guide the modulation process. It influences the direction and strength of the modulation applied during sampling. There are no specific minimum or maximum values, as it is typically a list or array of tags.

mod_w

The mod_w parameter controls the weight of the modulation applied during sampling. It affects how strongly the quality tags influence the output. The default value is often set to a moderate level, allowing for balanced modulation.

quality_neg

This parameter provides a negative baseline for the steering axis, decoupled from the CFG negative. It helps in refining the modulation process by providing a contrast to the quality tags. The default value is an empty string, which falls back to the CFG negative.

mod_start_layer

The mod_start_layer parameter specifies the first block to receive modulation during sampling. It determines where the modulation process begins within the model's architecture. The minimum value is typically 0, with the default set to 8.

mod_end_layer

This parameter specifies the last block to receive modulation, exclusive. It determines where the modulation process ends within the model's architecture. The default value is set to 27, with -1 indicating the use of all available blocks.

mod_taper

The mod_taper parameter controls the number of late slots within the modulation range that are scaled by the taper_scale. It allows for gradual modulation towards the end of the process. The default value is 0, indicating no tapering.

mod_taper_scale

This parameter specifies the multiplier applied to the tapered slots during modulation. It affects the strength of the modulation in the tapered region. The default value is 0.25, allowing for subtle tapering.

mod_final_w

The mod_final_w parameter controls the weight of the modulation applied at the final layer. It determines the strength of the modulation at the end of the process. The default value is 0.0, indicating no additional modulation.

denoise

This parameter specifies the level of denoising applied during sampling. It affects the clarity and quality of the output. The default value is 1.0, indicating full denoising.

window_size

The window_size parameter controls the size of the window used for spectral forecasting. It affects the efficiency and quality of the sampling process. The default value is 2.0, allowing for balanced forecasting.

flex_window

This parameter specifies the flexibility of the window size during spectral forecasting. It allows for dynamic adjustment of the window size based on the sampling process. The default value is 0.25, providing moderate flexibility.

warmup_steps

The warmup_steps parameter specifies the number of initial steps used for warming up the sampling process. It helps in stabilizing the process before full sampling begins. The default value is 6, providing a balanced warmup period.

blend_w

This parameter controls the blending weight applied during the sampling process. It affects the smoothness and quality of the output. The default value is 0.3, allowing for moderate blending.

cheby_degree

The cheby_degree parameter specifies the degree of the Chebyshev polynomial used for feature forecasting. It affects the accuracy and efficiency of the forecasting process. The default value is 3, providing a balanced degree.

ridge_lambda

This parameter specifies the regularization strength for ridge regression used in the sampling process. It affects the stability and quality of the output. The default value is 0.1, providing moderate regularization.

dcw_mode

The dcw_mode parameter controls the mode of the dynamic contrast weighting applied during sampling. It affects the contrast and quality of the output. The default value is "off", indicating no dynamic contrast weighting.

dcw_lambda

This parameter specifies the regularization strength for dynamic contrast weighting. It affects the stability and quality of the output. The default value is 0.01, providing subtle regularization.

dcw_band_mask

The dcw_band_mask parameter specifies the band mask used for dynamic contrast weighting. It affects the regions of the output that receive contrast weighting. The default value is "LL", indicating a low-low band mask.

dcw_calibrator

This parameter specifies the calibrator used for dynamic contrast weighting. It affects the accuracy and quality of the contrast weighting process. The default value is AUTO_CALIBRATOR_SENTINEL, indicating automatic calibration.

cfgpp_lambda

The cfgpp_lambda parameter specifies the regularization strength for the CFG++ process. It affects the stability and quality of the output. The default value is 0.0, indicating no additional regularization.

fsg

The fsg parameter controls the use of frequency spectrum guidance during sampling. It affects the quality and efficiency of the output. The default value is False, indicating no frequency spectrum guidance.

fsg_band_lo

This parameter specifies the lower bound of the frequency band used for spectrum guidance. It affects the range of frequencies considered during sampling. The default value is 0.59, providing a balanced lower bound.

fsg_band_hi

The fsg_band_hi parameter specifies the upper bound of the frequency band used for spectrum guidance. It affects the range of frequencies considered during sampling. The default value is 0.75, providing a balanced upper bound.

fsg_k

This parameter specifies the number of frequency bands used for spectrum guidance. It affects the granularity and quality of the guidance process. The default value is 3, providing a balanced number of bands.

fsg_d_sigma

The fsg_d_sigma parameter specifies the standard deviation of the frequency bands used for spectrum guidance. It affects the spread and quality of the guidance process. The default value is 0.1, providing moderate spread.

fsg_gamma

This parameter specifies the gamma correction applied during frequency spectrum guidance. It affects the contrast and quality of the output. The default value is 0.0, indicating no gamma correction.

adaptive_smc_alpha

The adaptive_smc_alpha parameter controls the alpha value used for adaptive SMC-CFG during sampling. It affects the detail recovery and quality of the output. The default value is _SMC_CFG_ALPHA_DEFAULT, providing balanced adaptation.

smc_cfg_lambda

This parameter specifies the regularization strength for SMC-CFG during sampling. It affects the stability and quality of the output. The default value is _SMC_CFG_LAMBDA_DEFAULT, providing moderate regularization.