Random Integer in Range (Legacy)

Somewhere in almost every workflow there's a "roll a random number between these bounds and feed it to something" moment - a random strength for a controlnet, a random LoRA weight, a random pick from a set. ComfyUI core gives you seeds that randomize, but a plain integer roll that you can replay takes a bit more. Random Integer in Range (Legacy) is that: minimum, maximum, seed in, one integer out, fully deterministic for a given seed.

The mechanism is the point, so let's be honest about it: it's random.Random(seed).randint(min, max), nothing more. Python's random module is deterministic given the same seed, which means the node is reproducible run-to-run - the same seed always yields the same integer. That's what makes it useful in a way that a raw seed widget isn't: you can fix a seed, get a value, and come back to it later without having memorized the number.

The inputs that matter

• minimum / maximum - the bounds. It doesn't care which is bigger; it swaps them internally, so 1..10 and 10..1 both work. Full signed 64-bit range, so you can roll something huge if you need it.
• seed - the reproducibility knob. This one has control_after_generate attached, just like a KSampler seed, so it can randomize on its own if you leave it. Lock it to fixed when you want the value to stay put. The standard ComfyUI footgun applies: if you leave the control on randomize, the shown seed is the next one, not the one that produced your current result.

The output

A single random_integer (INT). Wire it into any converted integer input - a scale factor, a strength, an index into a list - or into another node that consumes primitives. Because it's a proper INT socket, it fans out to as many consumers as you like without the "one value typed in five places" drift.

Installing it

This one lives in silveroxides/ComfyUI-UtilsCollection. ComfyUI Manager → search ComfyUI-UtilsCollection → install → restart, or clone it by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/silveroxides/ComfyUI-UtilsCollection

Restart after cloning. Requirements are just opencv-python and typing-extensions - nothing model-related, no downloads. The "(Legacy)" suffix again means this is a compatibility alias for the pack's canonical UC_IntegerRangeRandom; you may see the newer name in the menu.

Where people get burned

The reproducible-seed thing is the main trap in reverse: people assume any "random" node is non-deterministic and don't realize this one replays. If you're deliberately rolling a fresh value every run, make sure the seed's control_after_generate is set to randomize - with it on fixed, you'll quietly get the same integer forever, which is either exactly what you wanted or a subtle bug depending on your workflow. And since randint is inclusive on both ends, 1..10 can return 10, which trips people up when they're using the output as a list index and their list has 9 items.