BMAB KSampler

This is the ordinary first-generation step in a BMAB pipeline - the node that turns whatever your Integrator packaged (model, clip, vae, prompt, seed) into an actual first image. It's the plainest sampler in the pack, and the one tell that confirms it: denoise defaults to 1, meaning a full generation from noise, not a touch-up pass. Compare that to BMAB KSampler Hires. Fix (denoise defaults to 0.4) or any of the detailer nodes (denoise around 0.4-0.45) - across this whole pack, the denoise default is basically a label telling you whether a node is meant to generate or to refine. This one generates.

What it needs

steps, cfg_scale, sampler_name, scheduler - the standard set. The sampler and scheduler dropdowns default to "Use same sampler" / "Use same scheduler," which on this node specifically means inheriting whatever's set on the bind coming out of your Integrator - worth setting your real choice here (or on the Integrator, if BMAB's Integrator carries one) since this is usually the first place in the graph those settings actually get used. denoise at its default of 1 is what you want for a standard txt2img-style generation; you'd only lower it here if you're deliberately doing img2img off a latent or image fed in through the Integrator.

The only optional input is lora - BMAB's own lora type, for applying a LoRA specifically to this generation pass.

Output is the standard pair: an updated bind to keep chaining into detailer and resize nodes downstream, and the generated image.

Where it sits in a typical graph

BMAB Integrator (or Flux Integrator) → BMAB KSampler → whatever refinement you want next: BMAB KSampler Hires. Fix for a resolution boost, one or more detailer nodes for faces/hands/people, BMAB Basic for a final color grade, BMAB Save Image to finish. This node is almost always the first thing that actually produces pixels in that chain.

Installing it

ComfyUI Manager: search comfyui_bmab, install, restart. By hand:

cd ComfyUI/custom_nodes
git clone https://github.com/portu-sim/comfyui_bmab.git
cd comfyui_bmab
pip install -r requirements.txt

plus comfyui_controlnet_aux and ComfyUI_IPAdapter_plus, the two companion packs the README lists as needed for comfyui_bmab as a whole, even though this specific node doesn't call either directly.

Where people get confused

The most common mix-up is reaching for this node when you actually want the hires-fix variant, or vice versa - they look nearly identical in the graph (same steps/cfg/sampler/scheduler/denoise shape) and the only functional signal that separates "generate" from "refine" is that one default denoise value. If you're chaining two sampler nodes back to back and the second one seems to be regenerating the image from scratch instead of adding detail, check you didn't accidentally drop a second BMAB KSampler where you meant BMAB KSampler Hires. Fix.