Nodes/advanced-cfg-controller/⚙️ Advanced CFG Controller (Expert)
ComfyUI Node

⚙️ Advanced CFG Controller (Expert)

A ComfyUI node in model_patches/Advanced_CFG_Controller with 53 inputs and 2 outputs.

By FlareAI-Studios·Created 7 months ago·Updated 7 months ago· 1
⚙️ Advanced CFG Controller (Expert)
  • model
  • attention_modifiers_positive
  • attention_modifiers_negative
  • attention_modifiers_fake_negative
  • attention_modifiers_global
  • MODEL
  • STRING
adaptive_cfgHard Intensity
adaptive_cfg_topk0.25
adaptive_cfg_ref8.0
skip_uncondtrue
uncond_sigma_start1000.0
uncond_sigma_end1.0
fake_uncond_startfalse
fake_uncond_expfalse
fake_uncond_exp_methodUse Positive
fake_uncond_exp_value2.0
fake_uncond_exp_normalizefalse
fake_uncond_multiplier1
fake_uncond_sigma_start1000.0
fake_uncond_sigma_end1.0
lerp_uncondfalse
lerp_uncond_strength2.0
lerp_uncond_sigma_start1000.0
lerp_uncond_sigma_end1.0
subtract_latent_meanfalse
subtract_latent_mean_sigma_start1000.0
subtract_latent_mean_sigma_end1.0
latent_intensity_rescalefalse
latent_intensity_rescale_methodHard Measurement
latent_intensity_rescale_cfg8.0
latent_intensity_rescale_sigma_start1000.0
latent_intensity_rescale_sigma_end3.0
cond_expfalse
cond_exp_normalizefalse
cond_exp_sigma_start1000.0
cond_exp_sigma_end1.0
cond_exp_methodAmplify
cond_exp_value2.0
uncond_expfalse
uncond_exp_normalizefalse
uncond_exp_sigma_start1000.0
uncond_exp_sigma_end1.0
uncond_exp_methodAmplify
uncond_exp_value2.0
attention_modifiers_global_enabledfalse
disable_condfalse
disable_cond_sigma_start1000.0
disable_cond_sigma_end0.0
enable_performance_monitoringfalse
adaptive_noise_strength0.00
spectral_norm_target1.0
frequency_filter_cutoff0.50
dynamic_range_compression1.00
args_filter
Categorymodel_patches/Advanced_CFG_Controller

Inputs (53)

NameTypeDefaultDescription
modelMODEL
adaptive_cfgCOMBOHard Intensity6 options: Disabled, Soft Intensity, Hard Intensity, Hard Squared, Range Based, Spectral Analysis
adaptive_cfg_topkFLOAT0.250–0.5Top-K percentage for intensity measurement (lower = more precise)
adaptive_cfg_refFLOAT8.00–100Reference CFG value for adaptive scaling
skip_uncondBOOLEANtrueSkip negative prompt calculation for speed boost
uncond_sigma_startFLOAT1000.00–10000Start skipping negative at this sigma level
uncond_sigma_endFLOAT1.00–10000Stop skipping negative at this sigma level
fake_uncond_startBOOLEANfalseUse fake negative prompt when skipping
fake_uncond_expBOOLEANfalseApply experimental modifications to fake negative
fake_uncond_exp_methodCOMBOUse PositiveMethod for generating fake negative prompt
fake_uncond_exp_valueFLOAT2.00–1000Strength value for fake negative method
fake_uncond_exp_normalizeBOOLEANfalseNormalize fake negative to preserve magnitude
fake_uncond_multiplierINT1-1–1Multiplier for fake negative (-1, 0, or 1)
fake_uncond_sigma_startFLOAT1000.00–10000Start using fake negative at this sigma
fake_uncond_sigma_endFLOAT1.00–10000Stop using fake negative at this sigma
lerp_uncondBOOLEANfalseBlend positive and negative prompts
lerp_uncond_strengthFLOAT2.00–10Blending strength (higher = more negative influence)
lerp_uncond_sigma_startFLOAT1000.00–10000Start blending at this sigma level
lerp_uncond_sigma_endFLOAT1.00–10000Stop blending at this sigma level
subtract_latent_meanBOOLEANfalseFix color drift by centering latent values
subtract_latent_mean_sigma_startFLOAT1000.00–10000Start color correction at this sigma
subtract_latent_mean_sigma_endFLOAT1.00–10000Stop color correction at this sigma
latent_intensity_rescaleBOOLEANfalseRescale latent intensity to prevent oversaturation
latent_intensity_rescale_methodCOMBOHard MeasurementMethod for measuring latent intensity
latent_intensity_rescale_cfgFLOAT8.00–100Target intensity level for rescaling
latent_intensity_rescale_sigma_startFLOAT1000.00–10000Start rescaling at this sigma
latent_intensity_rescale_sigma_endFLOAT3.00–10000Stop rescaling at this sigma
cond_expBOOLEANfalseApply experimental modifications to positive prompt
cond_exp_normalizeBOOLEANfalseNormalize positive prompt to preserve magnitude
cond_exp_sigma_startFLOAT1000.00–10000Start modifications at this sigma
cond_exp_sigma_endFLOAT1.00–10000Stop modifications at this sigma
cond_exp_methodCOMBOAmplifyModification method for positive prompt
cond_exp_valueFLOAT2.00–100Strength value for positive modification
uncond_expBOOLEANfalseApply experimental modifications to negative prompt
uncond_exp_normalizeBOOLEANfalseNormalize negative prompt to preserve magnitude
uncond_exp_sigma_startFLOAT1000.00–10000Start modifications at this sigma
uncond_exp_sigma_endFLOAT1.00–10000Stop modifications at this sigma
uncond_exp_methodCOMBOAmplifyModification method for negative prompt
uncond_exp_valueFLOAT2.00–100Strength value for negative modification
attention_modifiers_global_enabledBOOLEANfalseEnable global attention modifications
disable_condBOOLEANfalseDisable positive prompt in specified range
disable_cond_sigma_startFLOAT1000.00–10000Start disabling positive at this sigma
disable_cond_sigma_endFLOAT0.00–10000Stop disabling positive at this sigma
enable_performance_monitoringBOOLEANfalseTrack performance metrics during generation
adaptive_noise_strengthFLOAT0.000–1Strength of adaptive noise injection
spectral_norm_targetFLOAT1.00.1–5Target norm for spectral normalization
frequency_filter_cutoffFLOAT0.500.1–1Cutoff frequency for frequency domain filtering
dynamic_range_compressionFLOAT1.000.1–1Dynamic range compression ratio (1.0 = disabled)
args_filteroptSTRINGComma-separated list of parameters to include in output
attention_modifiers_positiveoptATTNMODAttention modifications for positive prompt
attention_modifiers_negativeoptATTNMODAttention modifications for negative prompt
attention_modifiers_fake_negativeoptATTNMODAttention modifications for fake negative
attention_modifiers_globaloptATTNMODGlobal attention modifications

Outputs (2)

NameTypeDescription
MODELMODEL
STRINGSTRING