Integer

Komfy_Integer is a plain integer value node: type a number, get an INT out the other end as a wire. Default 1024, which is a hint - the SDXL generation size - and the range spans the full signed 32-bit window, from about −2.1 billion to +2.1 billion.

The wide range isn't a flex, it's practical. Integers in ComfyUI do triple duty. The obvious jobs are seeds (which can legitimately be any 32-bit value) and generation dimensions like 1024. But they're also steps, batch sizes, frame counts for video, and index numbers for anything that picks from a list. A node that accepts the whole range means you're never stuck because the widget won't take the value you need.

How it works

One required input, value (INT, default 1024, min −2147483648, max 2147483647, step 1) and one output, value, a typed INT socket. It's a straight passthrough, re-executing every run so changes flow immediately.

The pattern that makes it useful is the same one the other Komfy nodes lean on: wire it into a KSampler seed, a width or height input, a step counter - anywhere an integer widget lives - and you get a single authoritative source you can fan out to several nodes. "One seed, five samplers" works because the seed is defined once and wired everywhere; the same idea applies to any number you want to keep consistent. Retitle the node (Seed, Width, Frames) and the graph reads itself.

Install

Part of the HMT Suite pack:

cd ComfyUI/custom_nodes
git clone https://github.com/tuyenhm68/ComfyUI-HMT-Suite.git
cd ComfyUI-HMT-Suite
pip install -r requirements.txt

Restart ComfyUI, then find it under HMT Suite > Komfy > Integer. ComfyUI Manager also works (search "ComfyUI-HMT-Suite").

The honest take

ComfyUI ships PrimitiveInteger, so you're not getting functionality the core lacks - you're getting the HMT pack's version with the full 32-bit range and the label-from-title convention. If you already have the pack installed for the downloader or updater nodes, this slots in as a clean way to wire seeds and sizes. If you're choosing between this and the core primitive, there's no wrong answer; pick whichever sits in your palette. The one thing to remember: an integer is only as useful as the node it's wired into, and here it's wired into everything that takes a whole number.