Math Operation

Math Operation is the node you reach for when a workflow needs a bit of arithmetic and you can't be bothered wiring up three separate add/multiply/divide nodes. It takes two numbers, A and B, applies one of five operations - +, -, *, /, ^ - and hands you the result twice: once as an integer and once as a float.

That dual output is the quietly smart part. ComfyUI's sockets are picky - some nodes want an INT, others want a FLOAT - so a single node that emits both means you never have to add a convert node just to satisfy a socket type. It's the same value underneath, just presented in two wrappers.

How it works

Nothing clever going on. A and B come in as floats, the operation is chosen from a dropdown, and the code does exactly what you'd expect. Two details worth knowing:

• ^ is exponentiation (A ** B), not XOR or bitwise anything. If you want power-of-2 dimensions, 2 ^ 10 gives you 1024.
• Division by zero doesn't silently give you inf - it raises ZeroDivisionError, which shows up as a red node and stops the queue. If a divide is going to see a zero upstream, guard it or pick a different operation.
• The integer output uses Python's int(), which truncates toward zero. int(7.9) is 7, and int(-7.9) is -7. Good to remember if you're feeding the result into something like a resolution or a batch count where rounding direction matters.

Inputs and outputs

The whole interface:

• A (FLOAT) - left operand.
• B (FLOAT) - right operand.
• operation - the five-way dropdown.

Outputs, always both:

• result_int (INT)
• result_float (FLOAT)

Where you'd actually use it

The boring-sounding answer is "everywhere." Need steps = 20 * (1 - denoise)? That's A=20, B=denoise, operation=- with a multiply upstream - or just compute it with - if you're feeling fancy about the ordering. Scaling a width for a hires pass, deriving a batch count from a seed range, computing an aspect-corrected dimension from a base number - this node covers the one-off math that otherwise gets you hunting through the primitive menu. It's not going to replace a real expression node for anything with three or more operands; that's what chaining this with itself is for.

Installing it

It ships in ComfyUI Fictiverse Nodes. ComfyUI Manager, search "ComfyUI Fictiverse Nodes", install, restart. Or the manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/Fictiverse/ComfyUI_Fictiverse

No dependencies, no models, pure Python math. The pack is Apache 2.0.

Common issues

Same pack-level gotcha as the rest of Fictiverse: if the folder isn't named exactly ComfyUI_Fictiverse, the node won't load and you'll see ModuleNotFoundError: No module named 'custom_nodes.ComfyUI_Fictiverse' in the console. Rename the folder and restart.

The one node-specific trap is dividing by zero and wondering why your queue silently fails. The error message names the node, but it's easy to miss if you have a chain of math feeding a sampler. If a denominator can be zero - say, a value that starts at 0 - clamp it upstream or add a tiny epsilon via +.