Configurable Float Slider

The Configurable Float Slider is the one you reach for when you want to sweep a value instead of typing into a box. ComfyUI's default number widgets work, but dragging the little label is fiddly and there's no visual sense of where you are in the range. This node replaces that with a proper draggable slider and a big live readout - the same idea as a Primitive value node (one authoritative source you can fan out to several sockets, per the plumbing docs in the KB) but actually pleasant to touch.

It's the flagship of the ComfyUI-SimpleFloatSlider pack by Carasibana. The pack is tiny, MIT-licensed, and ships exactly three nodes under utils/sliders; this is the fully-configurable float one. Wire its output into a converted widget - LoRA strength on a LoRA loader, denoise, CFG, ControlNet weight - and you can scrub the value while you iterate without touching the target node. It saves the same value it displayed, which is more than some slider nodes manage.

How it works

There's a split personality here, and it's the right kind. The Python node is dumb on purpose: it just takes the five inputs, clamps value into [min_value, max_value], and rounds to precision decimals. All the drag, wheel, and click-to-type behavior lives in a JavaScript frontend widget that ships with the pack - no server round-trip, so dragging never re-queues a run. The number you see is the number the graph uses when you hit Queue.

The inputs that matter

The collapsed node shows just value (a slider, default 0.5). Hit the ▾ configure toggle and the real controls appear:

• min_value / max_value - the range, defaults 0.0–1.0. This is why the node exists; set it to whatever your target actually accepts.
• precision - decimal places shown, 0–4, default 2. Display only; the rounded value is what comes out.
• step - how far dragging or scrolling moves, default 0.01. Note it's independent of precision: you can show two decimals while stepping in 0.25 chunks.

One output, float, holding the current value clamped and rounded. Convert a target node's widget to input and wire this into it.

The stepping gotcha (worth reading once)

The README's rule of thumb: you sit on the grid unless you deliberately type your way off it. If you type 37 on a slider stepping by 5, you get exactly 37, and dragging then goes 42, 47 - it doesn't snap back onto the 35/40 grid. Editing the step field re-grids from min_value. Clamping happens after stepping, so min and max are always reachable even when the grid would overshoot. If you ever wonder why a value looks off-grid, that's it - retype the step and it tidies up.

Installing it

Zero dependencies, zero model files, zero config. Via ComfyUI Manager, search "ComfyUI-SimpleFloatSlider" and install; or clone it in:

cd ComfyUI/custom_nodes
git clone https://github.com/Carasibana/ComfyUI-SimpleFloatSlider

Restart ComfyUI and all three nodes show up under utils/sliders. Because there's no requirements.txt, there's nothing to go wrong - the whole pack is one Python file plus one JS widget.

Troubleshooting

• Slider renders as a plain widget - restart ComfyUI fully after installing; the JS frontend only loads on startup.
• Saved range ignored on load - that was a real v1.1.1 bug (workflows reloaded with default bounds); it's fixed. If you're on an old version, update the pack.
• Value looks clamped wrong - narrowing the range clamps the display, but raising it back restores your original value since 1.1.1, so nothing is destroyed by a quick experiment.

For most float tuning, this is the node. If you only ever need 0–1, the pack's Simple Float Slider is the zero-config version; this one is for everything else.