Merge sigmas by average

When no single scheduler does what you want

Every scheduler in ComfyUI is a curve with a personality. Karras is middle-heavy, simple is flat, exponential front-loads the work. Merge sigmas by average from sigmas_tools_and_the_golden_scheduler lets you split the difference: feed it two SIGMAS tensors and it returns a weighted average of them, so you get a schedule that's, say, 60% Karras and 40% simple. It's the musical-chords version of schedulers - two existing sounds combined into one.

The math is a one-liner under the hood:

sigmas_1 * proportion_1 + sigmas_2 * (1 - proportion_1)

Every step gets the same blend weight. That's the whole design: constant, predictable, no surprises.

What you set

Three inputs:

• sigmas_1 and sigmas_2 - the two schedules you're blending
• proportion_1 - a float from 0 to 1 (default 0.5) controlling how much of sigmas_1 survives. 0.5 is a straight 50/50; 0.8 leans hard toward schedule 1; 0 leans entirely to schedule 2

One SIGMAS output, wired into a KSamplerAdvanced's sigmas input like any schedule.

The honest take

This is the node to reach for when you've found that one scheduler is almost right - you love Karras's structure but wish the finish were gentler, or you like exponential's start but want a calmer middle. Blend a couple of steps' worth in the right direction and you can often get exactly the curve you were hunting for across the whole dropdown. It's also a great way to learn what a scheduler actually contributes: blend simple with Karras, graph the result, and the shape tells you everything.

One practical constraint: both inputs must be the same length, since the average is element-wise. Feed it a 20-step and a 30-step schedule and you'll get an error, not a resized one. If that bites, this is the pack - it also ships Merge sigmas gradually (same inputs, but the blend weight ramps across the steps instead of staying flat), which is the version to use when you want to start with one scheduler and end with another.

Install & gotchas

ComfyUI Manager → search sigmas_tools_and_the_golden_scheduler, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Extraltodeus/sigmas_tools_and_the_golden_scheduler

Restart, no model downloads. If the pack errors on import it's the usual dependency gap - requirements.txt lists only asteval while the code imports matplotlib and scipy, so pip install matplotlib scipy asteval fixes a minimal install.