EasyUse/PreSamplinggenerated

PreSampling (NoiseIn)(samplerSettingsNoiseIn)

PreSampling (NoiseIn) (class samplerSettingsNoiseIn) is a pipeline pre-sampling node in the EasyUse/PreSampling category.

sampler Settings Noise In

pipe
optional_latent
pipe
factor
0.10
steps
20
cfg
8
sampler_name
COMBO
scheduler
COMBO
denoise
1.00
seed
INT
optional_noise_seed
INT
Easy Use

PreSampling (NoiseIn) (class samplerSettingsNoiseIn) is a pipeline pre-sampling node in the EasyUse/PreSampling category. It prepares sampling settings for a downstream sampler and adds a controllable noise-injection stage to the latent. It takes one PIPE_LINE, applies the configured sampler settings, optionally injects noise into a latent, and returns the updated pipeline. It does not execute sampling itself; feed its output into a KSampler or another node that consumes a PIPE_LINE.

Description

This node is used before a sampler to override the sampling behavior carried in the pipeline. In addition to the usual steps, cfg, sampler_name, scheduler, denoise, and seed settings, it provides a noise-injection path controlled by factor. This lets you add a controlled amount of noise to the latent before the sampler runs, with optional separate controls for the noise seed and the target latent.

The node does not produce a sampled image. It outputs an updated PIPE_LINE containing the sampling settings and any supplied noise/latent information.

Inputs

Required

InputTypeDefaultDescription
pipePIPE_LINEThe pipeline carrying the model, conditioning, latent, and existing sampler settings.
factorFLOAT0.1Strength of the noise injected into the latent before sampling. Range 0.0 to 1.0. 0.0 adds no noise; 1.0 applies full-strength noise.
stepsINT20Number of sampling steps to use. Range 1 to 10000.
cfgFLOAT8Classifier-free guidance scale. Range 0 to 100.
sampler_nameCOMBOSampler algorithm to use. Select from the available ComfyUI sampler names.
schedulerCOMBONoise schedule / scheduler to use for the sampler.
denoiseFLOAT1Amount of denoising to perform. Range 0.0 to 1.0. Lower values preserve more of the original latent content.
seedINTSeed for the sampling operation.

Optional

InputTypeDefaultDescription
optional_noise_seedINTA separate seed used for the noise-injection stage. Connect this when you want to control the injected noise independently from the main sampling seed.
optional_latentLATENTAn optional latent to use as the target for noise injection. If omitted, the latent already stored in the pipeline is used.

Outputs

OutputTypeDescription
pipePIPE_LINEThe updated pipeline with the configured sampling settings and optional noise-injection inputs applied. Pass this into a sampler node that accepts a PIPE_LINE.

Usage Notes

  • Place this node before the actual sampler node in a PIPE_LINE workflow.
  • Use factor to control how much noise is mixed into the initial latent. Small values produce subtle variation; higher values push the latent closer to pure noise.
  • denoise and factor work together: denoise controls how much of the latent is regenerated while factor controls how much noise is introduced before sampling starts.
  • Connect optional_noise_seed when you need to reproduce or vary the injected noise without changing the main sampler seed.
  • Connect optional_latent when you want to inject noise into a specific latent rather than the one stored in the pipe, such as an encoded image or a custom latent prepared upstream.
  • The returned pipe preserves the other contents of the original pipeline; only sampling settings and noise-related values are updated.

Comments

Sign in with GitHub to join the discussion.

Loading comments…