Scheduler Selector

The scheduler is the quietest of the sampler settings - everyone remembers sampler names, almost nobody remembers whether they used beta or karras last week. This node exists so the scheduler and step count live in one place, as data you can wire into other nodes, instead of being a dropdown you can't reach.

What it does

Two inputs: steps (default 20, range 1–10000) and scheduler_name, a dropdown populated with ComfyUI's real scheduler list (default beta). Two outputs: steps and scheduler, both plain values - the steps pass through validated, the scheduler comes out as a STRING. That's the whole node. It doesn't run anything; it just centralizes two values and hands them to anything that wants them.

Why you'd bother

It's the same job as the pack's Sampler Selector, but for the scheduler, with the step count thrown in. If you're wiring sampler info into a Construct Metadata node, or you have a workflow where steps need to be computed or passed around as data rather than typed into every sampler, this keeps the source of truth in one node. You pick the scheduler once, wire the string where it's needed, and the dropdown can't drift out of sync with what your metadata claims you used.

Worth noting: this outputs a plain string and an int, not a SAMPLER_INFO bundle - so it's for feeding text/int consumers, not for plugging straight into the pack's "KSampler w/ Sampler Info" node. If that's the goal, use Sage_SamplerInfo instead, which bundles the whole config.

Installing it

It ships in the Sage Utils pack. ComfyUI Manager (search Sage Utils) or:

cd ComfyUI/custom_nodes
git clone https://github.com/arcum42/ComfyUI_SageUtils
cd ComfyUI_SageUtils
pip install -r requirements.txt

Restart ComfyUI. The pack's only pip dependency is dynamicprompts; no model downloads needed.

Watch out for

The beta default is worth a moment's thought. beta is ComfyUI's default scheduler and it's a fine all-rounder, but if you're porting settings from an A1111 workflow, you're probably used to karras - the dropdown won't tell you which one you meant, so set it deliberately. And as with any scheduler choice: on distilled models, matching the scheduler to the model's training recipe matters more than personal preference.