Float Multi-Switch

Floats are the quiet majority of what you tune in ComfyUI - denoise strength, ControlNet weight, guidance, blend amounts. The Float Multi-Switch gives you five FLOAT inputs and passes through the first one that isn't None. It's the same "first live input wins" pattern as the pack's other multi-switches, and it exists for the same reason: so you can offer a menu of preset values and let bypass do the choosing.

How to use it

Wire preset values into the slots - say a denoise of 0.3, 0.5, 0.7 - and feed the output to your KSampler's denoise port. To pick one, you don't touch the node at all: you bypass the presets you don't want, and the switch hands through whatever's left live. It's an option menu made entirely of ComfyUI's native bypass, which means it plays nicely with group bypassers and workflow organizers.

The ordering rule from the author applies here too: put the things you'll disable on the first slots, and anything that's always live on the last slot - because an always-connected slot-1 input is never None, so the switch will stop there and never reach the others. If you want slot 4 to win, slots 1–3 must be genuinely empty (bypassed upstream).

There's a catch specific to floats, though: all five inputs are forceInput, meaning you can't type a value directly into the node. The numbers have to come from upstream nodes - the pack's own Float primitive, another switch, a math node, whatever. So it's not a place to punch in a quick number; it's a router for values that already exist elsewhere in the graph. That's the trade-off for being wire-driven instead of widget-driven.

What's on the node

• input1–input5 - optional FLOAT inputs, wire-only.
• float output - one FLOAT, ready for denoise, ControlNet weight, or anything else that eats a float.

If every slot is empty, it doesn't return a soft default - it raises a hard "Missing Input" error. The author's philosophy is loud errors over silent garbage, and for a value that feeds into sampling, that's the right call.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2
# restart ComfyUI

Or via ComfyUI Manager, searching "ComfyUI-RvTools_v2". No models, no downloads; the pack's deps are the standard torch/numpy/Pillow/opencv-python/pilgram set.

The caveat

Like everything in this pack: the README banners it as unmaintained and recommends ComfyUI_Eclipse as the successor. The original RvTools repo was also yanked from GitHub in early 2025, breaking shared workflows and sparking the usual "where do I get this now" thread on r/comfyui. v2 is the renamed, still-installable version - fine for existing workflows, and Eclipse is where new builds should look.