Float Multiply

Scaling is the most common thing you do to a number in a graph. Take a value, multiply it by some factor, feed the result on. JWFloatMul is exactly that and nothing more: two floats a and b in, a * b out.

The reason a whole node exists for one multiply sign is the same reason ComfyUI exists at all - you want the math in the pipeline, not scribbled on a sticky note. When the factor is itself wired from somewhere else, the multiplication recomputes on every run. Bump the upstream value and everything downstream that depends on this product moves with it. One source of truth beats five hardcoded copies of the same guess.

What it actually does

a times b, both FLOAT, one FLOAT out. Both default to 0 (so a fresh node multiplies to zero until you set something), and the range is enormous (±1e17) - treat it as unbounded for any real use.

Typical jobs: scaling a resolution by a factor before a resize, turning a base strength into a scaled one (multiply a LoRA weight by 0.5 for a subtle pass), converting a normalized 0–1 slider into an actual pixel count, or halving a value by multiplying by 0.5. Any "this, but scaled" is a JWFloatMul.

The inputs and outputs that matter

Two required inputs, one output - that's the entire surface:

• a - first factor (FLOAT).
• b - second factor (FLOAT).

Output: a single FLOAT equal to a * b. Wire it into a resize factor, a denoise, a CFG, or the next node in a math chain.

How to install it

Grab the pack jamesWalker55/comfyui-various. Via ComfyUI Manager, search Various ComfyUI Nodes by Type, install, restart. Or clone it:

cd ComfyUI/custom_nodes
git clone https://github.com/jamesWalker55/comfyui-various

and restart. No dependencies - the primitive math nodes are plain Python. Worth knowing: the author built this pack so you can install just the file you need. All the Float and Integer nodes live in a single comfyui_primitive_ops.py; drop that one file into custom_nodes/ and you get the math without dragging in the image, mask, and audio nodes you didn't ask for. In an ecosystem where custom-node clutter is a genuine complaint, that's a nice touch.

Common issues & troubleshooting

The default is 0, so an untouched input zeroes your result. If JWFloatMul is spitting out 0, one of the inputs is still at its default. Sounds obvious; catches everyone once.

Consider whether you actually want a chain of these. Multiply, then add, then divide - three JW nodes and three wires. A single math-expression node (the kind newer packs ship, where you type a * b + c) collapses that into one box. If your graph is turning into a spreadsheet, switch. For a single clean multiply that you never want to think about again, JWFloatMul is the right amount of tool.