Sigmas Cleanup

If you're going to use exactly one node from RES4LYF's sigma-math toolbox as a safety net, make it this one. Sigmas Cleanup enforces a minimum noise floor across a schedule - a simple job, but a genuinely useful one once you've been running sigmas through the pack's other math nodes.

Why a floor matters

The very bottom of a noise schedule, where sigma approaches zero, is where samplers get numerically fragile. Values that are too close to zero - or that dip below zero entirely, which several of this pack's other sigma nodes can produce as a side effect - tend to cause NaNs, wasted steps, or outright instability rather than useful extra denoising. A floor prevents any value in the schedule from going below a set minimum.

The default value is a real number, not a placeholder

sigmin defaults to 0.0291675 - an oddly specific figure rather than a round number like 0.01 or 0.05. That specificity is a good sign: it reads like a value tuned to an actual minimum sigma a real model uses, not an arbitrary default someone typed in. Worth knowing that if you're working with a different model family, this default may not be the right floor for you - but it's clearly not arbitrary either.

How it works

Every value in the input schedule that falls below sigmin gets clamped up to it (or removed, depending on the implementation - the README doesn't spell it out, and this specific node isn't documented anywhere). Either way, the practical effect is the same: nothing in the output schedule ends up below the floor.

The inputs and outputs that matter

• sigmas (SIGMAS, required) - the schedule to clean up.
• sigmin (default 0.0291675, range 0–1000) - the minimum allowed value.

Output is a single SIGMAS list.

When to actually use it

Right before your sampler, after any hand-editing or math-node chain - Sigmas2 Add, Sigmas2 Mult, the arc-trig reshaping nodes, the chaos and attractor generators all covered elsewhere in this pack are exactly the kind of operations that can leave degenerate near-zero or negative values at the tail of a schedule. Sigmas Cleanup is the node you add as a habit, not the one you reach for after something's already gone wrong.

How to install it

• ComfyUI Manager - search "RES4LYF", install, restart.
• Manual - activate your venv, cd ComfyUI/custom_nodes && git clone https://github.com/ClownsharkBatwing/RES4LYF, cd RES4LYF, pip install -r requirements.txt (portable builds: use the embedded pip.exe). Restart.

Common issues & troubleshooting

Getting odd artifacts or NaNs after building a custom sigma schedule? This is the first node to add, not the last resort - put it directly before your sampler in the chain.

Setting sigmin too high can visibly change your output. It's a floor, so raising it well above the default effectively shortens how far the denoising process is allowed to go - you'll lose the very finest detail the last few steps would otherwise resolve. If your final images look slightly under-refined after adding this node, that's the tradeoff; lower sigmin back toward the default (or your model's actual minimum) if it's costing you more than it's fixing.

It doesn't fix ordering, only the floor. If your schedule is jagged or non-monotonic further up the curve, Sigmas Cleanup won't smooth that out - pair it with Sigmas CatmullRom if the whole shape needs work, not just the bottom.