Nodes/WAS Affine/Ultimate Affine KSampler [Latent]
ComfyUI Node

Ultimate Affine KSampler [Latent]

A ComfyUI node in latent/sampling with 24 inputs and 1 output.

By WASasquatch·Created 12 months ago·Updated 4 months ago· 5
Ultimate Affine KSampler [Latent]
  • latents
  • model
  • positive
  • negative
  • affine_schedule
  • external_mask
  • options
  • noise_options
  • LATENT
seed0
steps20
cfg8.0
sampler_nameeuler
schedulersimple
denoise0.20
affine_interval1
max_scale1.2000
max_bias0.0000
patternwhite_noise
affine_seed0
affine_seed_incrementfalse
tile_width0
tile_height0
tile_overlap8
batch_size0
Categorylatent/sampling

Inputs (24)

NameTypeDefaultDescription
latentsLATENTInput latents dict with key 'samples'. Supports [B,C,H,W] (images) or [B,C,F,H,W] (video).
modelMODELDiffusion model to sample with.
positiveCONDITIONINGPositive prompt conditioning.
negativeCONDITIONINGNegative prompt conditioning.
seedINT00–18446744073709550000Seed for base sampler (noise).
stepsINT201–10000Number of denoising steps.
cfgFLOAT8.00–100Classifier-free guidance scale. Can be a single float value or a per-step list. Short lists repeat last value.
sampler_nameCOMBOeulerBase sampler algorithm.
schedulerCOMBOsimpleScheduler for sigma schedule.
denoiseFLOAT0.200–1Denoise fraction (<=1).
affine_intervalINT11–100Apply affine every N steps (1 = every step).
max_scaleFLOAT1.20000–2Scale at schedule peak: 1 + (max_scale-1)*t.
max_biasFLOAT0.0000-2–2Bias at schedule peak: max_bias*t.
patternCOMBOwhite_noiseAffine mask pattern.
affine_seedINT00–18446744073709550000Seed for affine mask generation.
affine_seed_incrementBOOLEANfalseIncrement affine seed after each application (temporal).
affine_scheduleDICTWASAffineScheduleOptions dict.
external_maskoptIMAGEOptional external mask to gate affine (adapted to latent space internally).
optionsoptDICTBase options for Affine (common/full options).
noise_optionsoptDICTPattern-specific overrides layered onto 'options'.
tile_widthoptINT0-1–2048Latent-space sampling tile width. 0 disables tiling (single pass). -1 auto-calculates proportional tile size ≤128 latent pixels (1024px image space with 8x compression).
tile_heightoptINT0-1–2048Latent-space sampling tile height. 0 disables tiling (single pass). -1 auto-calculates proportional tile size ≤128 latent pixels (1024px image space with 8x compression).
tile_overlapoptINT80–256Latent-space overlap between sampling tiles.
batch_sizeoptINT00–4096Process the batch dimension in chunks of this size to reduce peak VRAM. 0 = process all at once.

Outputs (1)

NameTypeDescription
LATENTLATENT