LC Sampler Configure Simple

Most sampler config nodes these days ship with a dual-pass split as a selling point, and most people never need it. If you run a single CFG through a plain sampler - which is honestly where most of us live, especially with the distilled models that want one number and nothing fancy - LC Sampler Configure Simple is the one to grab. It's the full LC Sampler Configure with the step_swap and cfg_2 machinery ripped out.

The pitch is dead simple: one node that owns steps, denoise, one CFG, the sampler, and the scheduler, and hands them back as sockets you can fan out. No second pass, no handoff index, no "which CFG applies to which half." If you don't know whether you need dual-pass sampling, you don't need it.

The inputs

• total_steps - full schedule length (default 40).
• detailer_steps - steps reserved for a detailer/refiner stage; 0 = unused.
• denoise - strength, 1.0 = full.
• cfg - the one guidance scale. Set it for your architecture: 5–9 on SD 1.5/SDXL and the anime finetunes, ~1 on guidance-distilled models (Klein, Z-Image Turbo, ERNIE Turbo). On those, CFG 7 isn't "more prompt adherence," it's double-applied guidance that burns the image at twice the compute - the KB's CFG-scale panel covers this at length.
• sampler_name (44 choices) and scheduler (9 choices) - the standard dropdowns. On flow-matching models, Euler + a conservative schedule (normal, beta, sgm_uniform) is the move; DPM++ 2M Karras remains the SDXL-family default.

Outputs mirror the inputs: total_steps, cfg, denoise, sampler_name, scheduler, detailer_steps - six typed sockets ready to wire into a KSampler or a KSamplerAdvanced.

Install

It's in the LC123 pack - Manager (search "LC123") or:

cd ComfyUI/custom_nodes
git clone https://github.com/lonecatone23/ComfyUI_LC123_nodes

Restart. The pack adds no pip dependencies beyond ComfyUI's own torch/numpy and is MIT licensed. Cloud-hosted ComfyUI (RunningHub and the like) won't have LC123 preinstalled - request it or run local, or every LC node in a shared workflow throws a missing-node error.

The take

This is the one I'd reach for as a default. It does the central-config job without dragging you into two-pass architecture you didn't ask for. If a workflow later genuinely needs a step swap and a second CFG, you swap this for the full LC Sampler Configure - the sockets line up closely enough that it's a two-minute change.