Nodes/KarmaNodes/Karma KSampler Cycle
ComfyUI Node

Karma KSampler Cycle

A ComfyUI node in KarmaNodes/Sampling with 40 inputs and 2 outputs.

By KarmaSwint·Created about a year ago·Updated 5 months ago· 1
Karma KSampler Cycle
  • primary_model
  • vae
  • latent_image
  • positive
  • negative
  • secondary_model
  • upscale_model
  • pos_additive
  • neg_additive
  • latent(s)
  • vae
steps20
seed0
cfg8.00
sampler_name
scheduler
starting_denoise1.00
cycle_denoise0.50
total_cycles2
enable_denoise_scalingtrue
denoise_min_threshold0.10
enable_steps_scalingfalse
steps_scaling_value5
steps_adjustment_modedecrement
steps_threshold_modeauto
steps_manual_threshold5
secondary_model_start_cycle2
upscale_factor2.0
upscale_methodbasic
scale_sampling
latent_upscale_methodlanczos
enable_gradual_upscalingfalse
gradual_upscale_steps3
pos_add_strength0.25
enable_pos_strength_scalingfalse
pos_add_strength_cutoff2.00
neg_add_strength0.25
enable_neg_strength_scalingfalse
neg_add_strength_cutoff2.00
sharpen_strength0.00
sharpen_radius2
use_tiled_vaefalse
CategoryKarmaNodes/Sampling

Inputs (40)

NameTypeDefaultDescription
primary_modelMODEL
vaeVAE
latent_imageLATENT
positiveCONDITIONINGThe main positive conditioning.
negativeCONDITIONINGThe main negative conditioning.
stepsINT201–10000
seedINT00–18446744073709550000
cfgFLOAT8.000–100
sampler_nameCOMBO44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38
schedulerCOMBO9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3
starting_denoiseFLOAT1.000–1
cycle_denoiseFLOAT0.500–1
total_cyclesINT22–12
enable_denoise_scalingBOOLEANtrueEnable automatic denoise strength scaling between cycles
denoise_min_thresholdFLOAT0.100.01–1Minimum denoise threshold to prevent going too low
enable_steps_scalingBOOLEANfalseEnable automatic steps adjustment between cycles
steps_scaling_valueINT51–50Amount to adjust steps by each cycle
steps_adjustment_modeCOMBOdecrementWhether to decrease or increase steps each cycle
steps_threshold_modeCOMBOautoAuto calculates threshold based on initial steps, manual uses fixed value
steps_manual_thresholdINT51–200Manual threshold value (only used when threshold mode is manual)
secondary_model_start_cycleINT21–12Cycle at which to switch to secondary model (if provided)
upscale_factorFLOAT2.00.1–8
upscale_methodCOMBObasicbasic = PIL image upscale, model = upscale model, latent = direct latent space resize (fastest, no VAE decode/encode)
scale_samplingCOMBOResampling method for image upscaling
latent_upscale_methodCOMBOlanczosResampling method for latent space upscaling
enable_gradual_upscalingBOOLEANfalseEnable gradual upscaling with multiple intermediate steps
gradual_upscale_stepsINT31–10Number of gradual upscaling steps (only used when gradual upscaling is enabled)
pos_add_strengthFLOAT0.250.01–1
enable_pos_strength_scalingBOOLEANfalseEnable positive strength scaling between cycles
pos_add_strength_cutoffFLOAT2.000.01–10
neg_add_strengthFLOAT0.250.01–1
enable_neg_strength_scalingBOOLEANfalseEnable negative strength scaling between cycles
neg_add_strength_cutoffFLOAT2.000.01–10
sharpen_strengthFLOAT0.000–10
sharpen_radiusINT21–12
use_tiled_vaeBOOLEANfalseUse tiled VAE processing for large images
secondary_modeloptMODELOptional secondary model to switch to at specified cycle
upscale_modeloptUPSCALE_MODEL
pos_additiveoptCONDITIONINGAdditional positive conditioning to blend in.
neg_additiveoptCONDITIONINGAdditional negative conditioning to blend in.

Outputs (2)

NameTypeDescription
latent(s)LATENT
vaeVAE