Nodes/RES4LYF/Sigmas Power
ComfyUI Node Runs on cloud

Sigmas Power

Reshape a schedule with a single exponent

By ClownsharkBatwing·Created 2 years ago·Updated 18 days ago· 1,222
Sigmas Power
  • sigmas
  • SIGMAS
power1.00

Sigmas Power is the most minimal sigma-shaping node in RES4LYF: it takes a SIGMAS schedule and raises it to a power. One float, one effect. Bending a noise curve with a single exponent is the crudest possible way to redistribute denoising effort across steps - but it's also the fastest to reason about, and sometimes that's exactly what you want when a fancier reshaper is overkill.

Raising sigmas to a power above 1 pushes the curve one way (concentrating change), below 1 pushes it the other. It's the same intuition as a gamma curve on an image: a single knob that biases where the action happens.

How it works

The node applies sigma ^ power element-wise across your incoming schedule and hands back the modified curve. Because it's a pure transform on any SIGMAS input, it chains after a scheduler like ClownScheduler or a stock one - you build a base schedule, then bias it with a power.

The inputs and output

  • sigmas (SIGMAS) - the schedule to reshape. Required; this node doesn't generate a schedule, it modifies one.
  • power (default 1) - the exponent. At 1 the curve is unchanged (identity). Above 1 and below 1 bend it in opposite directions. Because the effect compounds across the whole curve, small moves matter - nudge it in small steps rather than jumping to extremes.

The single output is SIGMAS → a sampler's sigmas input, or another sigma node in a chain.

Installing it

ComfyUI Manager: search RES4LYF, install, restart. Manually:

cd custom_nodes
git clone https://github.com/ClownsharkBatwing/RES4LYF/
cd RES4LYF
pip install -r requirements.txt

pip install inside the venv (or portable Python), restart, hard-refresh F5. No downloads.

Common issues

The honest truth: for most people, Sigmas Easing is the more useful reshaper - it gives you named curves and control over which end bends, where this gives you one blunt exponent. Reach for Power when you want a simple, monotonic bias and nothing more. The failure mode is treating it like a magic quality dial: too large a power warps the schedule into something the sampler handles badly, and you'll see it as artifacts or washed-out detail. Default is 1 (no change) for a reason. Move it a little, preview with SigmasPreview, and back off if the curve looks wrong. And, as always, it needs a schedule wired into its sigmas input - it can't conjure one. If the node's missing, RES4LYF isn't installed.

CategoryRES4LYF/sigmas

Inputs (2)

NameTypeDefaultDescription
sigmasSIGMAS
powerFLOAT1.00-100–100

Outputs (1)

NameTypeDescription
SIGMASSIGMAS