Nodes/MD Nodes/MD: PingPong FBG (Full Control)
ComfyUI Node

MD: PingPong FBG (Full Control)

A ComfyUI node in MD_Nodes/Samplers with 57 inputs and 1 output.

By MDMAchine·Created about a year ago·Updated 2 months ago· 14
MD: PingPong FBG (Full Control)
  • scheduler
  • SAMPLER
step_random_modeblock
step_size4
seed80085
first_ancestral_step0
last_ancestral_step-1
ancestral_noise_typegaussian
start_sigma_index0
end_sigma_index-1
enable_clamp_outputfalse
blend_modelerp
step_blend_modelerp
fbg_sampler_modeEULER
cfg_scale1.00
cfg_start_sigma1.00
cfg_end_sigma0.00
fbg_start_sigma1.00
fbg_end_sigma0.00
max_guidance_scale10.00
initial_guidance_scale1.00
fbg_guidance_multiplier1.00
guidance_max_change1000.00
pi0.50
t_00.50
t_10.40
fbg_temp0.00
fbg_offset0.00
fbg_eta0.00
fbg_s_noise1.00
max_posterior_scale3.00
log_posterior_initial_value0.00
log_posterior_ema_factor0.00
adaptive_noise_scalingfalse
noise_scale_factor1.00
progressive_blend_modefalse
conditional_blend_modefalse
conditional_blend_sigma_threshold0.50
conditional_blend_function_nameslerp
conditional_blend_on_changefalse
conditional_blend_change_threshold0.10
clamp_noise_normfalse
max_noise_norm1.00
gradient_norm_trackingfalse
enable_profilingfalse
debug_mode0
tensor_memory_optimizationfalse
early_exit_threshold0.00
ancestral_start_sigma1.00
ancestral_end_sigma0.00
sigma_range_presetCustom
enable_restartsfalse
yaml_settings_str
checkpoint_steps_str
restart_modebalanced
restart_noise_scale0.50
restart_s_noise1.00
restart_steps
CategoryMD_Nodes/Samplers

Inputs (57)

NameTypeDefaultDescription
step_random_modeCOMBOblockSTEP RANDOMIZATION MODE • Purpose: Controls how the seed evolves during sampling steps. • Options: off (static), block (periodic), reset (per-step), step (incremental). • Trade-offs: 'reset' is chaotic, 'off' is rigid. ⭐ Recommended: 'block' for best texture variation.
step_sizeINT41–100STEP SIZE INTERVAL • Purpose: Defines how many sampling steps occur before the seed rotates. ⭐ Recommended: 4-10 steps.
seedINT800850–9007199254740991RANDOM SEED • Purpose: Base JS-Safe seed for noise generation.
first_ancestral_stepINT0FIRST ANCESTRAL STEP • Purpose: The step index to begin injecting ancestral noise. ⭐ Recommended: Increase to 3-5 if initial structural composition is messy.
last_ancestral_stepINT-1LAST ANCESTRAL STEP • Purpose: The step index to stop injecting ancestral noise. ⭐ Recommended: -1 (Disable entirely) for standard deterministic diffusion.
ancestral_noise_typeCOMBOgaussianANCESTRAL NOISE TYPE • Purpose: The statistical distribution of the injected noise. • Options: Gaussian (standard bell), Uniform (flat limits), Brownian (correlated). ⭐ Recommended: Brownian for highly organic/analog textures.
start_sigma_indexINT0START SIGMA INDEX • Purpose: Skip early sampling steps (used for img2img routing).
end_sigma_indexINT-1END SIGMA INDEX • Purpose: Halt sampling early.
enable_clamp_outputBOOLEANfalseCLAMP OUTPUT • Purpose: Forces all output latent values strictly into the [-1, 1] range. ⭐ Recommended: False for audio generation (audio latents often exceed 1.0).
schedulerSCHEDULERSCHEDULER INPUT • Purpose: Connects the sigma schedule (from Basic/Hybrid/GITS node).
blend_modeCOMBOlerpBLEND MODE (PRIMARY) • Purpose: Mathematical approach for mixing new noise with the latent. ⭐ Recommended: 'lerp' for speed, 'slerp' for spherical structural integrity.
step_blend_modeCOMBOlerpSTEP BLEND MODE (SECONDARY) • Purpose: Interpolation used for internal dual-step calculations.
fbg_sampler_modeCOMBOEULERFBG INTERNAL MODE • Purpose: Defines the fundamental ODE solver step logic. ⭐ Recommended: EULER for standard Ace-Step audio models.
cfg_scaleFLOAT1.00-1000–1000CFG SCALE (BASE) • Purpose: Classifier-Free Guidance magnitude.
cfg_start_sigmaFLOAT1.00CFG START SIGMA
cfg_end_sigmaFLOAT0.00CFG END SIGMA
fbg_start_sigmaFLOAT1.00FBG START SIGMA
fbg_end_sigmaFLOAT0.00FBG END SIGMA
max_guidance_scaleFLOAT10.00MAX GUIDANCE CAP • Purpose: Hard ceiling limit for dynamic CFG scaling.
initial_guidance_scaleFLOAT1.00INITIAL GUIDANCE • Purpose: Starting CFG magnitude.
fbg_guidance_multiplierFLOAT1.00FBG MULTIPLIER • Purpose: Strength of the feedback loop correction.
guidance_max_changeFLOAT1000.00MAX CHANGE PER STEP • Purpose: Velocity limit for guidance scaling.
piFLOAT0.50PI (POSTERIOR FACTOR)
t_0FLOAT0.50T0 OFFSET
t_1FLOAT0.40T1 TEMP FACTOR
fbg_tempFLOAT0.00FBG TEMPERATURE
fbg_offsetFLOAT0.00FBG OFFSET
fbg_etaFLOAT0.00DEPRECATED: Legacy ETA parameter.
fbg_s_noiseFLOAT1.00DEPRECATED: Legacy noise scale.
max_posterior_scaleFLOAT3.00MAX POSTERIOR SCALE
log_posterior_initial_valueFLOAT0.00INITIAL LOG POSTERIOR
log_posterior_ema_factorFLOAT0.00POSTERIOR EMA
adaptive_noise_scalingBOOLEANfalseADAPTIVE NOISE SCALING
noise_scale_factorFLOAT1.00NOISE SCALE FACTOR
progressive_blend_modeBOOLEANfalsePROGRESSIVE BLEND
conditional_blend_modeBOOLEANfalseCONDITIONAL BLEND
conditional_blend_sigma_thresholdFLOAT0.50BLEND THRESHOLD
conditional_blend_function_nameCOMBOslerpCONDITIONAL BLEND FUNC
conditional_blend_on_changeBOOLEANfalseBLEND ON CHANGE
conditional_blend_change_thresholdFLOAT0.10CHANGE THRESHOLD
clamp_noise_normBOOLEANfalseCLAMP NOISE NORM
max_noise_normFLOAT1.00MAX NOISE NORM
gradient_norm_trackingBOOLEANfalseGRADIENT TRACKING: Unused.
enable_profilingBOOLEANfalseENABLE PROFILING
debug_modeINT00–2DEBUG MODE
tensor_memory_optimizationBOOLEANfalseMEMORY OPTIMIZATION
early_exit_thresholdFLOAT0.00EARLY EXIT THRESHOLD
ancestral_start_sigmaFLOAT1.00ANCESTRAL START
ancestral_end_sigmaFLOAT0.00ANCESTRAL END
sigma_range_presetCOMBOCustomSIGMA RANGE PRESET
enable_restartsBOOLEANfalseENABLE RES 2 RESTARTS
yaml_settings_stroptSTRINGYAML OVERRIDE
checkpoint_steps_stroptSTRINGCHECKPOINT STEPS
restart_modeoptCOMBObalancedRESTART MODE
restart_noise_scaleoptFLOAT0.50RESTART NOISE SCALE
restart_s_noiseoptFLOAT1.00RESTART S_NOISE
restart_stepsoptSTRINGCUSTOM RESTART STEPS

Outputs (1)

NameTypeDescription
SAMPLERSAMPLER