What It Does
KSamplerAdvanced provides the same core sampling functionality as KSampler but adds explicit control over the noise injection step and the start/end step range. This enables techniques like partial denoising (img2img), multi-pass generation (sampling different step ranges with different models), and hi-res fix workflows.
The add_noise parameter controls whether fresh noise is injected before sampling begins. When set to "disable", the node assumes the input latent already has the appropriate noise level—essential for chained sampling passes.
Return_with_leftover_noise controls whether the output latent retains residual noise, which is useful for feeding into subsequent sampling stages.
Inputs
modelMODELThe loaded diffusion model.
positiveCONDITIONINGPositive prompt conditioning.
negativeCONDITIONINGNegative prompt conditioning.
latent_imageLATENTInput latent.
add_noiseSTRINGWhether to add noise before sampling (enable/disable).
noise_seedINTRandom seed for noise generation.
stepsINTTotal number of steps.
cfgFLOATCFG scale.
sampler_nameSTRINGSampling algorithm.
schedulerSTRINGNoise schedule.
start_at_stepINTStep to begin sampling at.
end_at_stepINTStep to stop sampling at.
return_with_leftover_noiseSTRINGWhether to keep residual noise in output.
Outputs
LATENTLATENTDenoised latent image.
What Numonic Captures
- All KSampler parameters plus start/end step range
- Noise injection mode
- Multi-pass step configuration
Known Gaps
- Relationship between chained sampling passes — each pass is recorded independently
- Effective denoise ratio from step range — must be computed from start/end/total
Related Nodes
Capture ComfyUI metadata automatically
Numonic extracts workflow metadata from every ComfyUI generation — models, samplers, seeds, prompts, and custom nodes. Track provenance, maintain compliance, and never lose a workflow.