Sigmas GaussianCDF

Sigmas GaussianCDF is the focused, single-purpose sibling of Sigmas Gaussian. Where that node makes you pick from five different Gaussian-based operations, this one just does one of them - the cumulative distribution function - with two parameters instead of five. If you already know you want an S-curve-shaped reshaping of your schedule and don't need the pack's broader Gaussian toolbox, this is the more direct route.

What the CDF actually does to a curve

A Gaussian's cumulative distribution function is the classic S-shape: it starts flat near zero, rises steeply through the middle (around mu), then flattens out again approaching 1. Running a sigma schedule through it concentrates the rate of change in the middle of the run and compresses it at both ends - conceptually adjacent to what a Karras-style schedule does by bunching denoising effort in the middle steps, just derived from a different piece of math.

How it works

Every value in sigmas gets passed through the Gaussian CDF parameterized by mu (the center) and sigma (the spread), producing a value in the 0–1 range, which normalize_output then rescales back into something usable as an actual sigma schedule.

The inputs and outputs that matter

• sigmas (SIGMAS, required) - the schedule to transform.
• mu (default 0, range -10 to 10) - where the S-curve is centered.
• sigma (default 1, range 0.01–10) - how wide the transition region is. Smaller values make the S steeper; larger values stretch it out.
• normalize_output (default true) - rescale the result back into a usable sigma range.

Output is a single SIGMAS list.

Should you use this?

There's no dedicated documentation for it, and I found nothing in community discussion pointing to a specific recipe. Because it's a narrower version of Sigmas Gaussian's cdf operation with the exact same underlying math, the honest answer is: use whichever one is more convenient for your graph. If you're only ever going to want the CDF shape and don't want to think about the operation dropdown, this is the cleaner node. If you might want to switch between pdf, cdf, and the others later, build with Sigmas Gaussian instead so you're not swapping nodes to change modes.

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

A very small sigma produces an almost step-function output - most values collapse toward either extreme of the range, with only a narrow band of the schedule actually transitioning. If your output looks like it's lost most of its gradation, widen sigma.

Keep normalize_output on. A raw Gaussian CDF only ever outputs values between 0 and 1, which is nowhere near the range a typical sigma schedule needs (peaking well above 1 for most models) - without normalization your sampler will get a schedule that's far too compressed to do meaningful denoising work.

No community-tested defaults exist. Treat mu/sigma as parameters to explore with a preview node, not settings with a known-good starting point.