Seed [LP]

A small node doing a small job, and the job is worth doing anyway: hold one seed value in exactly one place so every node that needs it - a sampler, a logging step, this pack's own Pipe nodes, which have a dedicated seed slot - pulls from the same source instead of you retyping the same number into three different widgets and inevitably letting them drift out of sync.

What it does

There's nothing clever here by design. You set a number, it comes out the other side as an INT, and anything downstream that takes an integer seed can wire straight into it. The value is up to you to control from outside - this node isn't running its own randomization logic, it's just a single, shareable source of truth for a number you'd otherwise be duplicating across your graph by hand.

Inputs and outputs that matter

• seed (required, INT, default 0, range 0 up to roughly 1.8 × 10¹⁹) - the value itself. That upper bound is standard-issue for a 64-bit unsigned seed field, the same range you'll see on most sampler seed widgets in ComfyUI.

Output: seed INT - wire it into a KSampler's seed input, into the seed slot on this pack's Pipe or Pipe In nodes if you're bundling it with the rest of your pipeline, or anywhere else that expects a plain integer.

Installing it

Through ComfyUI Manager, search "ComfyUI Level Pixel" or "ComfyUI-LevelPixel." Or:

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

Restart ComfyUI. Nothing to download beyond the pack itself - this node has no dependencies of its own.

Where people get burned

Honestly, not much to trip over with a node this small - the main thing worth flagging is a workflow-design habit rather than a bug. If you're duplicating the same literal seed number into several different sampler nodes by hand instead of routing them all through one Seed node, you're setting yourself up to change it in one place, forget the others, and spend ten minutes later wondering why two branches of your workflow are producing different results with what you thought was "the same seed." Route it through one node and that class of mistake stops being possible.