PreSampling (Custom)(samplerCustomSettings)

PreSampling (Custom) is an EasyUse/PreSampling pipeline node that lets you fully customize how a later sampling pass is run. It takes the pipe, applies the selected guider, CFG, sampler, scheduler, sigma schedule, denoise, and noise settings, then returns the same pipe with those settings embedded for the downstream sampler node.

Description

This node is a preprocessing step for custom sampling workflows. Instead of configuring a sampler directly, it stores all the custom sampling parameters in a PIPE_LINE, which is then consumed by the matching sampler node. This is useful when you want to use the same sampling configuration repeatedly, or when you need a level of control beyond what the standard KSampler nodes expose.

Inputs

Required

• pipe (PIPE_LINE)
  The pipeline object from a preceding EasyUse node. It should contain the model, conditioning, and usually a latent to sample from.

• guider (COMBO)
  The guidance mode to use when sampling. Options are:

  • Basic
  • CFG
  • DualCFG
  • IP2P+CFG
  • IP2P+DualCFG
  • IP2P+Basic

  Basic applies no standard negative CFG guidance. CFG uses standard classifier-free guidance. DualCFG uses both cfg and cfg_negative. The IP2P variants are InstructPix2Pix-style combinations for models or workflows that use that conditioning.

• cfg (FLOAT)
  The CFG scale for standard guidance. Default is 3.5, with a range from 0.0 to 100.0.

• cfg_negative (FLOAT)
  The negative-side CFG scale used by DualCFG and IP2P+DualCFG guiders. Default is 1.5, range 0.0 to 100.0.

• sampler_name (COMBO)
  The sampler to use, such as euler, dpmpp_2m, etc. Ignored if an optional_sampler is provided.

• scheduler (COMBO)
  The scheduler used to generate the sigma schedule. Ignored if optional_sigmas is provided.

• coeff (FLOAT)
  A coefficient for guider modes that support a scaling coefficient. Default is 1.2, range 0.8 to 1.5, step 0.05.

• steps (INT)
  Number of sampling steps. Default is 20, range 1 to 10000.

• sigma_max (FLOAT)
  Starting noise level for generated sigma schedules. Default is 14.614642, range 0 to 1000.

• sigma_min (FLOAT)
  Ending noise level for generated sigma schedules. Default is 0.0291675, range 0 to 1000.

• rho (FLOAT)
  Schedule curvature parameter used by rho-based schedules such as Karras-style generation. Default is 7, range 0 to 100.

• beta_d (FLOAT)
  Beta schedule parameter for schedulers that use beta-based sigma generation. Default is 19.9, range 0 to 1000.

• beta_min (FLOAT)
  Minimum beta value for beta-based schedulers. Default is 0.1, range 0 to 1000.

• eps_s (FLOAT)
  Small epsilon term used by beta-based schedulers. Default is 0.001, range 0 to 1, step 0.0001.

• flip_sigmas (BOOLEAN)
  Reverses the sigma schedule when enabled. Default is false.

• denoise (FLOAT)
  Amount of denoising to apply. Default is 1, range 0 to 1, step 0.01. A value of 1.0 denoises fully; lower values preserve more of the initial latent.

• add_noise (COMBO)
  Controls whether and how noise is added at the start of sampling. Default is enable (CPU). Options are:

  • enable (CPU) — add noise using the CPU RNG.
  • enable (GPU=A1111) — add noise using the GPU RNG, matching A1111 behavior.
  • disable — do not add initial noise.

• seed (INT)
  The random seed used for noise generation during sampling. No default; you must supply an integer.

Optional

• image_to_latent (IMAGE)
  An optional image that is encoded into a latent and used as the starting latent for the sampling pass.

• latent (LATENT)
  An optional latent that replaces the latent normally carried in the pipe for this sampling pass.

• optional_sampler (SAMPLER)
  A sampler object that overrides the sampler_name dropdown.

• optional_sigmas (SIGMAS)
  A custom sigma schedule that overrides the generated schedule from the scheduler settings.

Outputs

• pipe (PIPE_LINE)
  The same pipeline object with the custom sampling settings stored in it, ready for the downstream sampler node.