Sigmas Gaussian

Sigmas Gaussian applies Gaussian (normal-distribution, bell-curve) math to a sigma schedule, and it's more flexible than it sounds - the operation dropdown picks which piece of Gaussian math actually runs: the probability density, the cumulative distribution, its inverse, or a couple of blend modes. This is squarely in RES4LYF's curve-reshaping family, the same territory as beta57 and Sigmas GammaBeta, just built on the normal distribution instead of the beta distribution.

The five operations, briefly

• pdf - the bell curve itself. Values near mean get weighted up, values far from it weighted down. This concentrates the transform's effect around one point in the schedule.
• cdf - the cumulative version: an S-shaped curve that smoothly ramps from 0 to 1 as you move through the distribution. Applying this reshapes a schedule into that S-curve gradient rather than a peak.
• inverse_cdf - the mathematical inverse of the above; converts values back the other direction through the S-curve.
• transform (the default) - a general Gaussian-based reshaping, presumably combining the mean/std parameters into a direct value transform rather than a probability computation.
• modulate - likely uses the Gaussian as a weighting curve applied multiplicatively against the input rather than replacing values outright, similar in spirit to a windowing function.

None of these five are spelled out further by the pack itself, so that's a reasonable reading of standard Gaussian terminology rather than a guarantee of the exact implementation.

The inputs and outputs that matter

• sigmas (SIGMAS, required) - the schedule to transform.
• mean (default 0, range -10 to 10) - the center of the distribution.
• std (default 1, range 0.01–10) - the spread. Smaller means a tighter, more concentrated effect; larger spreads it out.
• operation (enum: pdf, cdf, inverse_cdf, transform, modulate; default transform) - which of the five modes above runs.
• normalize_output (default true) - rescale the result into a usable range afterward.

Output is a single SIGMAS list.

Should you use this?

If you want a Gaussian-CDF-specifically reshaped schedule with fewer knobs, Sigmas GaussianCDF is the dedicated, simpler version of one slice of what this node offers (its cdf operation). This broader node is worth reaching for once you already know which of the five operations you want and roughly what mean/std should be - otherwise it's five different math operations behind one dropdown with no worked examples anywhere in the pack or the community. I found no shared settings or writeups for this node specifically, which fits the general pattern: RES4LYF's named, documented tools (beta57, the ClownsharKSampler settings) have community-tested defaults; its deep sigma-math toolbox mostly doesn't.

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

Switching operation changes the character of the result completely - pdf and cdf in particular produce very differently-shaped curves from the same mean/std. If a setting that worked for one operation looks wrong after you swap modes, that's expected; re-tune mean and std for the new operation rather than assuming they carry over.

inverse_cdf is numerically touchy near the extremes (values very close to 0 or 1 in a CDF blow up toward infinity in its inverse) - if you get NaNs or extreme spikes, that's the likely cause. Keep normalize_output on and check the result with a preview before it reaches a sampler.

No community-tested starting values exist for this node. Treat the defaults as a place to start experimenting from, not a known-good setting.