Sigmas GammaBeta

The gamma and beta functions are the kind of thing you meet in a stats or special-functions course and rarely again - gamma generalizes the factorial to non-integers, beta is built from ratios of gammas and shows up constantly in probability distributions. Sigmas GammaBeta runs your noise schedule through one of five variants of these - plain gamma, plain beta, and the "incomplete" partial-integral versions of each, plus log-gamma - and hands back a reshaped schedule.

Why reshape a schedule with a special function at all

This sits in the same family as RES4LYF's other curve-shaping sigma nodes: Karras schedules concentrate denoising effort using a power-law curve, beta57 uses the beta distribution, and this node generalizes that idea to a wider menu of curve shapes drawn from special-function math rather than the handful of named schedulers ComfyUI ships. The pitch, implicitly, is more control over exactly how effort gets distributed across steps than a fixed scheduler dropdown offers - at the cost of picking values for functions most people haven't touched since a stats class.

The inputs and outputs that matter

• sigmas (SIGMAS, required) - the schedule to reshape.
• function (enum: gamma, beta, incomplete_gamma, incomplete_beta, log_gamma; default gamma) - which special function drives the transform.
• offset (default 0, range -10 to 10) and scale (default 0.1, range 0.01–10) - shift and scale applied around the function, standard knobs for repositioning a curve.
• parameter_a / parameter_b (both default 0.5, range 0.1–10) - the shape parameters that feed the chosen function. For beta/incomplete_beta these are the classic alpha/beta parameters (the same pair beta57 fixes at 0.5 and 0.7); for the gamma variants they likely control shape and rate, though the node doesn't document which does what for each function choice.
• normalize_output (default true) - rescale the result back into a usable sigma range after the transform, which you generally want left on.

Output is a single SIGMAS list.

Should you use this?

If you already know what you're doing with gamma and beta distributions, this gives you real, fine-grained control over a schedule's shape - more than the fixed beta57 preset offers, since you're free to pick your own alpha/beta instead of the pack's tuned 0.5/0.7. If you don't, I wouldn't start here. I found no community writeups or shared settings for this specific node, and the safer path to "a beta-distribution-shaped schedule" for flow-matching models is just using beta57 directly - it's the single most widely adopted piece of RES4LYF, with a known-good parameterization, while this node is the unguided general case behind it.

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

Small changes to parameter_a/parameter_b can swing the curve a lot, especially near the low end of their range - these functions aren't linear in their parameters. Change one at a time and watch the shape with a preview node before committing to a render.

incomplete_gamma/incomplete_beta involve numerical integration under the hood, which is inherently more sensitive to edge-case inputs than the closed-form gamma/beta options. If you get NaNs or a schedule that collapses to near-constant values, try the non-incomplete variant first to confirm the base shape looks right, then switch back.

Always keep normalize_output on unless you have a specific reason not to. These functions can produce wildly different output magnitudes depending on the parameters, and an un-normalized result can leave you with a schedule whose sigmas are nowhere near the range your sampler expects.