Nodes/ComfyUI-FlowMatching-Upscaler/Flow Matching Progressive Upscaler
ComfyUI Node

Flow Matching Progressive Upscaler

A ComfyUI node in latent/upscaling with 28 inputs and 5 outputs.

By ttulttul·Created 9 months ago·Updated 6 months ago· 62
Flow Matching Progressive Upscaler
  • model
  • positive
  • negative
  • latent
  • latent
  • next_seed
  • model
  • positive
  • negative
seed0
steps_per_stage16
cfg4.5
sampler_name
scheduler
total_scale4.00
stages2
renoise_start0.35
renoise_end0.15
skip_blend_start0.80
skip_blend_end0.05
upscale_methodbicubic
noise_schedule_override
skip_schedule_override
denoise1.00
enable_dilated_samplingenable
dilated_downscale2.00
dilated_blend0.25
dilated_min_steps1
dilated_seed_modederive
dilated_denoise0.50
cleanup_stagedisable
cleanup_noise0.00
cleanup_denoise0.40
Categorylatent/upscaling

Inputs (28)

NameTypeDefaultDescription
modelMODELFlow-matching diffusion model to drive refinement.
positiveCONDITIONINGPositive conditioning for CFG.
negativeCONDITIONINGNegative conditioning for CFG.
latentLATENTLow resolution latent to progressively upscale.
seedINT00–18446744073709550000Base seed controlling the re-noising at each stage.
steps_per_stageINT161–256Denoising steps executed at every progressive stage.
cfgFLOAT4.50–20Classifier Free Guidance strength per stage.
sampler_nameCOMBOSampler backend leveraged during refinement.
schedulerCOMBONoise schedule applied during denoising.
total_scaleFLOAT4.001–16Overall scale factor from the input latent to the final output.
stagesINT21–5Number of progressive stages to reach the total scale.
renoise_startFLOAT0.350–1Noise ratio applied at the first stage.
renoise_endFLOAT0.150–1Noise ratio applied at the last stage (interpolated in-between).
skip_blend_startFLOAT0.800–1Blend weight for the upsampled latent at the first stage.
skip_blend_endFLOAT0.050–1Blend weight for the upsampled latent at the last stage.
upscale_methodCOMBObicubicAlgorithm used when resizing the latents between stages.
noise_schedule_overrideoptSTRINGComma-separated override for per-stage noise ratios. Empty uses interpolated start/end.
skip_schedule_overrideoptSTRINGComma-separated override for per-stage skip blend weights.
denoiseoptFLOAT1.000–1Denoising strength supplied to the sampler per stage.
enable_dilated_samplingoptCOMBOenableOptionally run a dilated refinement pass for global coherence.
dilated_downscaleoptFLOAT2.001–4Factor used when downscaling for the dilated pass (>=1.0).
dilated_blendoptFLOAT0.250–1Blend weight of the dilated refinement result.
dilated_min_stepsoptINT11–256Minimum sampling steps for dilated refinement. Actual steps = max(min_steps, steps // 2). Increase for lightning models to ensure adequate sampling (e.g., 4).
dilated_seed_modeoptCOMBOderiveSeed mode for dilated sampling. 'derive' adds 10,000 to base seed, 'same' uses identical seed.
dilated_denoiseoptFLOAT0.500–1Denoising strength for dilated sampling. Lower values preserve more spatial structure from the original.
cleanup_stageoptCOMBOdisableRun an extra non-scaling clean-up denoise pass at the end.
cleanup_noiseoptFLOAT0.000–1Noise ratio for the optional clean-up stage (set to 0 to disable re-noising).
cleanup_denoiseoptFLOAT0.400–1Denoising strength used during the clean-up stage.

Outputs (5)

NameTypeDescription
latentLATENT
next_seedINT
modelMODEL
positiveCONDITIONING
negativeCONDITIONING