LC Seed Jump 🌱

LC Seed Jump is the most honest node in the pack: it does one arithmetic trick and doesn't pretend otherwise. You give it a seed a and a jump b, and it hands back six seeds - a, a+b, a+2b, all the way to a+5b. Six sockets, six stepped values, done.

Why would you want seeds spaced by a fixed step instead of just rolling six randoms? Controlled variation. Random seeds give you six arbitrary starting points; stepped seeds give you a structured sweep where each output is a predictable offset from the last. That's the difference between "show me six random images" and "show me six images I can reproduce by knowing the base and the jump." If you're iterating on a concept and want a grid of nearby variations - or you're building a batch that needs consistent spacing across multiple runs - this keeps the whole thing reproducible from two numbers.

How it works

The mechanism is in the name. seed is the base (input as a socket, so you can chain it from LC Seed or any INT source), jump is the step size with a hard range of 1–20. It computes a + i*b for i = 0 through 5 and emits them as seed_1 through seed_6. That's the entire node - no randomization, no modes, no hidden state. If you want the base itself to vary run-to-run, feed it from an LC Seed in randomize mode and the jump still spaces the six outputs.

The inputs

• seed - the base value a.
• jump - step size b, 1–20.

Outputs: seed_1 … seed_6, all INT.

Install

It's in the LC123 pack - Manager (search "LC123") or:

cd ComfyUI/custom_nodes
git clone https://github.com/lonecatone23/ComfyUI_LC123_nodes

Restart. No extra pip packages; MIT licensed. Cloud-hosted ComfyUI needs the pack requested or a local install.

The honest take: for most single-image work this is overkill - a plain seed with increment mode covers the "next one" case. Where Seed Jump earns its keep is batch/A-B graphs where you want several samplers on known-adjacent seeds at once. It's a tiny node, and that's fine; not everything in a 89-node pack needs to be a showstopper.