MD: Math Multiply (Int/Float)

Of the four basic math nodes in MD Nodes, multiply is the one you'll actually live in. Scaling a value is the most common operation in a parameterized ComfyUI workflow - double the steps, halve the denoise, convert a 0-to-1 control into a 0-to-10 range, or gate a strength by a global multiplier. MD_Math_Multiply does a × b and hands back both an int and a float result, which is the whole selling point.

You've probably already got a way to multiply in ComfyUI, so let's be honest about when this earns its slot: when you're tired of type-mismatch errors. The dual int_result / float_result outputs mean one node feeds whichever consumer you have - a steps input that wants an INT and a strength input that wants a FLOAT - without a converter in between.

How it works

Two FLOAT inputs, a and b. Result is a * b. The int output is the truncated product; the float output is exact. Inputs are full-range floats, so there's nothing to clamp or worry about - it's plain arithmetic in the wrapper.

The inputs that matter

• a - the base value being scaled.
• b - the multiplier. The tooltip's examples are the ones you'll actually use: 2.0 doubles, 0.5 halves. For global-strength scaling, connect a shared multiplier value here so one knob scales everything.

Outputs: int_result and float_result - take your pick.

Installing

It's in the MD Nodes pack:

cd path/to/ComfyUI/custom_nodes
git clone https://github.com/MDMAchine/ComfyUI_MD_Nodes.git
cd ComfyUI_MD_Nodes
pip install -r requirements.txt

Or ComfyUI Manager → MD Nodes → install → restart. Pure Python, no dependencies; it loads fine even if you skip the pack's requirements file (those matter for the audio and plotting nodes).

Common issues

Two things worth knowing. First, the int result truncates: 1.5 × 2.1 gives 3, not 4. If a downstream INT input is rounding-sensitive, feed it the float and let it convert, or the truncation will bite you. Second, don't over-reach on a single multiplier - if you're scaling several different kinds of parameters with one shared value, a b that's perfect for one range will be nonsense for another. The pack's LFO generator or a per-parameter multiply node is the better home for sweeping things; this node is for the steady, per-value scale.