CustomExponent

CustomExponent is PredefinedExponent's freer sibling: instead of picking e, 10, or 2 from a dropdown, you type any base you want and the node computes base^sigma for every value in your schedule. The base input starts at 2.0 with a minimum of 1.1, so you're always dealing with a genuinely exponential transform - nothing that degenerates into a flat line.

It's a niche utility, and it should be. The honest use case is the same as PredefinedExponent - taking a schedule out of log space after you've been editing it there, or deliberately reshaping a curve by exponentiating it - just with a continuous base instead of three presets. The extra freedom buys you finer control and costs you one more thing to reason about.

How it works

For each sigma value it computes base^sigma. Because the base is a continuous float, the shape of the transform is yours to tune: a base near 1.1 is nearly a no-op (1.1^small stays small), while a base of 4 or 5 turns modest input variation into dramatic output swings. There's no normalization, no clamping, no error handling for absurd results - the node computes what you asked for and hands it back.

The inputs and output

• sigmas - the schedule to transform.
• base (2.0, min 1.1, step 0.1) - the exponential base. Closer to 1.1 is gentler; higher is more aggressive.
• Output: SIGMAS - the transformed array.

Install

Part of ComfyUI-ScheduledGuider-Ext. ComfyUI Manager - search ComfyUI-ScheduledGuider-Ext - or:

cd ComfyUI/custom_nodes
git clone https://github.com/mfg637/ComfyUI-ScheduledGuider-Ext

Restart ComfyUI. No models, no extra dependencies.

Common issues

The same explosion trap as PredefinedExponent, squared: with a custom base you can overshoot even faster. Feed a real-world schedule through 5^x and you're generating 5-to-the-14th numbers for fun. This node is happiest on 0–1 curves or on schedules that genuinely live in log space; if the output looks like an astronomy constant, your input was in the wrong space.

Worth repeating the pack's wider rule of thumb here, because it applies to the whole exponent/logarithm family: these transforms are tools for working on schedules, not schedules themselves. Use them to reshape or round-trip a curve, then run the result through ScaleToRange before anything that actually drives a sampler.