Academia SD Numeric Input 🔢

Every ComfyUI canvas ends up with the same tiny annoyance: you type 1024 into a node, and the next node demands a FLOAT, or demands an INT, and you reach for a conversion node or manually retype it. This node is the two-for-one that kills the round-trip. Type a single integer once, and it emits both the INT and the FLOAT versions from one value - so the same number feeds a strict-INT input on one side and a FLOAT input on the other without a converter in between.

How it works

It's about as simple as a node gets. One required input, value (INT, range −99,999,999 to 99,999,999, step 1). The node casts it once: int_val = int(value) and float_val = float(value). Both outputs are always in sync because they come from the same widget - change the number once and both cables update together. No state, no caching, no hidden magic.

The two outputs are INT and FLOAT, both from the single value input. Wire the INT cable to things that demand integers (resolution inputs, batch sizes, a seed) and the FLOAT cable to anything that wants decimals (denoise, CFG, strength sliders that only accept FLOAT).

The real reason it exists

ComfyUI's type system is strict about input sockets - many nodes will refuse a mismatched type or silently coerce it in ways that surprise you. If a node keeps rejecting your number because it expects FLOAT, this is the clean fix: the pack's README positions it exactly as "avoid using additional converter nodes when connecting the same value to parameters that require strict data types." It's a small thing, but it's the kind of small thing that stops you from dragging a Convert Number To Float onto the canvas for the fifth time today.

Install

Pack install: ComfyUI Manager → "AcademiaSD", or git clone https://github.com/AcademiaSD/comfyui_AcademiaSD, restart. No dependencies.

Honest assessment

This is a utility, not a hero - you'll use it constantly and never talk about it. The one limitation worth knowing: the input is strictly an INT widget, so you can't type 2.5 and get a FLOAT out. For decimals you're still using a normal float input node. If that's all you need, plenty of utility packs (Impact, rgthree) offer combined converters too - but this one is already in the pack, zero-install, and it does the job without fuss.