SamplerPropertiesNode

SamplerPropertiesNode is a settings-packaging node: it takes the sampling parameters you'd normally scatter across your graph - checkpoint, steps, CFG, denoise, sampler, scheduler - and hands them back as six outputs you can route around. Nothing is sampled here; it's a single tidy source of truth for "how should this workflow sample", instead of six widgets in six places.

The use case is graph organization. If you have a workflow with several KSamplers or custom sampler nodes, you either duplicate the settings everywhere (and forget to update one of them) or you wire a shared settings node. This node is the shared-settings node. Set your defaults once, and every sampler downstream pulls from the same place. It also makes a workflow share-friendlier - a newcomer sees one "sampler properties" block instead of hunting for scattered sliders.

How it works

It's a config bundle, zero math, with one notable quirk.

Inputs:

• ckpt_name - the checkpoint, a dropdown populated from your installed checkpoints (the schema shows it as a dynamic list, which is why it's empty on paper - it fills at runtime).
• steps (INT, default 20) - sampling steps.
• cfg (FLOAT, default 8) - the CFG scale.
• denoise (FLOAT, default 8) - the denoise strength. Here's the quirk: the default is 8 in a field whose valid range is 0–1. That's a copy-paste slip in the pack - denoise must be between 0 and 1, so change it to something sane (0.5–0.8 for img2img, 1.0 for txt2img) before you run. This is exactly the kind of thing that silently breaks a workflow if you leave it.
• sampler_name - dropdown of the usual sampler list (euler, euler_ancestral, dpmpp_2m, ...).
• scheduler - dropdown (normal, karras, exponential, sgm_uniform, simple, ddim_uniform).

Outputs, in order: STRING (checkpoint name), INT (steps), FLOAT (cfg), FLOAT (denoise), STRING (sampler), STRING (scheduler). Route each to the matching input on your sampler node.

How to use it

Wire the outputs into a KSampler's steps, cfg, denoise, sampler_name, scheduler, and use the checkpoint name with a LoadCheckpoint. The payoff is when you have two samplers doing different jobs (an initial pass and a refiner, say) - you can set them both from one node and change one number instead of hunting. If you use a pack like Efficiency Nodes with its own all-in-one sampler, this is redundant; it shines on hand-built graphs.

Install and notes

Pack install, once: ComfyUI Manager → search "ComfyUI-LexTools", or git clone https://github.com/SOELexicon/ComfyUI-LexTools into custom_nodes, restart. No models, no downloads.

Two gotchas worth remembering. First, that denoise default of 8 - always set it explicitly. Second, the outputs are plain values, not a bundle type, so you still wire them one-by-one; the node's value is organizational, not a magic pipe. It's a tasteful convenience in a pack that mostly does heavy image analysis, and it tells you the author actually uses their own workflows.