Float

This is a node that holds a number. That's it. One float widget, one float output, and the entire job is giving that value somewhere to live on the canvas instead of buried in a widget on some other node. It sounds trivial until you've wired the same CFG value to five samplers and realized you want one authoritative place to change it.

This is the core "value node" pattern the KB's node-plumbing doc describes: pull a value out of the widget layer, make it a first-class object the graph can fan out from, and suddenly changing one number updates everywhere it's connected. ComfyUI core ships PrimitiveFloat for exactly this job, and FairLab's Float is the same idea wearing a simpler outfit - you type a value, it comes out the other side unchanged, and you can drag wires off it to as many consumers as you like.

The mechanism is a pass-through. Read the source and it's one line: return (value,). Input comes in as a widget, output goes out as a FLOAT, nothing else happens. The only real behavior of note is the widget itself: a step of 0.01, so the spinner moves in hundredths, and a default of 0.0.

Inputs and outputs:

value - the float you set on the widget, default 0.0, step 0.01.
value - the same float, now available as a wire.

Where it fits: anywhere you'd otherwise retype a number on several nodes, or where you want a value that downstream logic can read. It's the sort of node that looks pointless in a screenshot and becomes load-bearing the moment your graph grows past a dozen nodes. The same pack also ships Int and Number nodes for the integer and multi-output cases; grab whichever matches your data type.

One quirk worth knowing: despite being a value node, it's flagged OUTPUT_NODE = True in the source, and it's categorized under Fair/string rather than logic or utility. Neither matters for how you use it - it's just the pack author's bookkeeping - but don't go hunting for it under the wrong category in your node search.

Install is the standard pack path: ComfyUI Manager → search ComfyUI-FairLab → install → restart:

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

Restart, search "Float". No models, no dependencies, nothing to break - it's the most failure-proof node in this pack.

Inputs (1)

Outputs (1)