CFG Value (Umbra Lab)

This is a number in a box wearing a trench coat. CFG Value (Umbra Lab) exists so you can type your CFG scale once and fan it out to every sampler in the workflow, instead of editing the widget on each KSampler and silently forgetting one. That's the whole job, and it does it without pretending to be anything else.

Why a dedicated node beats the built-in widget

ComfyUI already lets you "convert widget to input" on any sampler's CFG field and wire in a Primitive node. The Umbra version just makes that one step less fiddly: drop CFG Value, set the number, and its cfg output is a FLOAT socket any sampler input can accept. The practical win is the one the whole plumbing layer exists for - repetition and illegibility. One graph with a base pass and a hires-fix pass that both need CFG 7 has two places to change the value; this node makes it one.

It's also handy in workflows that feed the value into a string or math path, because the output is a real typed FLOAT, so you can chain it into any node that consumes one without converting types by hand.

The inputs and output

It's almost absurdly thin, which is the point:

• cfg - the value itself, a FLOAT from 0.0 to 100.0, default 7.0, stepped in 0.1 increments. 7.0 is the classic SD1.5/SDXL starting point; modern distilled models like Flux and the Z-Image generation usually want 1.0 or below, so the range matters less than the habit of keeping it in one place.
• cfg (output) - the same FLOAT, ready to wire into a KSampler, the Umbra KSamplers, or any conditioning path that takes a float.

That's the whole schema. If you expected a randomize mode or a "value as string" option, you're in the wrong node - Umbra's Seed Value carries the randomization logic for the family, and this one is deliberately boring.

Installing it

It ships in Umbra-Nodes, the ComfyUI companion package for Umbra Studio (an open-source local AI creation suite by NocturneLabs). The easy route is ComfyUI Manager → Install Custom Nodes → search "Umbra Nodes". Or clone it directly:

cd ComfyUI/custom_nodes
git clone https://github.com/Nocturne-Ai-Labs/Umbra-Nodes

Restart ComfyUI afterward. The pack declares no pip dependencies, and this node touches nothing outside stock ComfyUI - it's just a FLOAT pass-through, so there's no model download and nothing to break.

Gotchas

The one thing that bites people with any value node is forgetting it's still just a widget: if you randomize the KSampler's own seed but leave a fixed CFG here, that's correct behavior - CFG isn't meant to randomize. And when you change the value here, remember ComfyUI re-runs everything downstream of it, so a graph that recomputes a slow model load is expected, not a bug. If you find yourself pasting the same number into five widgets anyway, this node is the cure you were looking for.