Float Math

Core ComfyUI is weirdly shy about basic arithmetic. You want to compute a strength as "1 minus the existing strength," or scale a width by 0.75, and there's no obvious node for it. Float Math is that node: two floats in, pick an operation, get the result - and it hands you both a float and an int version, so you rarely have to add a conversion step after it.

It's part of DebugPadawan's ComfyUI Essentials, a small MIT-licensed utility pack that fills exactly these boring gaps in the graph. No generation, no models - just glue that keeps your numbers flowing.

How it works

Under the hood it shares a calculation helper with the pack's Integer Math node. The operation set here is add, subtract, multiply, divide, power, max, min. One notable absence: no modulo - that's on the integer variant, not this one. If you need remainder math, use Integer Math instead.

The mechanism is worth one careful look because of what it doesn't do: division by zero doesn't error. The code guards it and returns 0.0. Same story for anything that would blow up - you get a zero, not a crash. That's convenient for robustness and a trap if you don't notice the math silently went to zero.

Inputs and outputs that matter

• a, b - the two operands, FLOAT inputs with 0.01 step on the widget.
• operation - the dropdown: add / subtract / multiply / divide / power / max / min.

Outputs are float_result (the real result) and int_result - which is int() of the float, meaning it truncates toward zero. 2.9 becomes 2, not 3. If you wanted rounding, this isn't it; that's what a Float to Int node with a round mode is for.

Install

Standard custom-node fare:

cd ComfyUI/custom_nodes
git clone https://github.com/DebugPadawan/DebugPadawans-ComfyUI-Essentials.git

Restart ComfyUI, or install via ComfyUI Manager by searching "DebugPadawan". requirements.txt lists numpy, torch, and opencv-python - numpy and torch already ship with ComfyUI and the shipped code never imports cv2, so no heavy downloads, no model files, no API keys.

Where people get burned

• The silent divide-by-zero. If your denominator can ever be zero (and in a graph it always can), you'll get 0.0 with no error to tell you. Clamp your divisors first.
• int_result truncates. For anything like CFG or strength math where rounding matters, grab the float and convert explicitly.
• No modulo here. People reach for this node expecting the full arithmetic kit; modulo lives on the integer version.