What It Does
KSampler is the heart of most ComfyUI workflows. It takes a loaded model, positive and negative conditioning (prompts encoded by CLIP), and a latent image, then iteratively denoises the latent space to produce an image.
The node exposes all critical generation parameters: seed (for reproducibility), steps (iteration count), CFG scale (prompt adherence strength), sampler name (euler, dpmpp_2m, etc.), scheduler (normal, karras, exponential), and denoise strength. These parameters directly control image quality and style.
For provenance, KSampler is the most important node to capture—its parameters are the primary determinants of the generated output alongside the prompt and model.
Inputs
modelMODELThe loaded diffusion model.
positiveCONDITIONINGPositive prompt conditioning from CLIPTextEncode.
negativeCONDITIONINGNegative prompt conditioning.
latent_imageLATENTInput latent (from EmptyLatentImage or VAEEncode).
seedINTRandom seed for reproducibility.
stepsINTNumber of denoising steps.
cfgFLOATClassifier-free guidance scale.
sampler_nameSTRINGSampling algorithm (euler, dpmpp_2m, etc.).
schedulerSTRINGNoise schedule (normal, karras, exponential).
denoiseFLOAToptionalDenoising strength (1.0 = full denoise).
Outputs
LATENTLATENTDenoised latent image.
What Numonic Captures
- Seed value (resolved, even when randomized)
- Steps count
- CFG scale
- Sampler name and scheduler
- Denoise strength
- Full node parameter snapshot in workflow JSON
Known Gaps
- Sampling duration — no timing data recorded
- Intermediate step outputs — only the final latent is preserved
- Memory usage during sampling
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.