Seed

The Seed node does one small, useful thing: it holds a seed value and outputs it, so you can define your seed once and feed it to several nodes instead of typing (and re-typing) the same number into each KSampler. If you've ever run a two-sampler workflow and forgotten to change the seed in the second one, you already understand why a dedicated seed source exists.

A seed is just the starting number for the random noise a sampler denoises from. Same seed + same everything = same image, every time. That reproducibility is the whole reason seeds matter, and pulling the seed out into its own node makes it something you control from one place.

How it works

You set a seed value on the node and it outputs that value. Wire the output into every node that takes a seed - your samplers, a WAS KSampler that accepts a seed input, a noise node - and now they all share one source of truth. Change it once and the whole graph moves together; you can't accidentally leave one sampler on a stale number.

That's the practical win over typing seeds inline. It's not doing anything clever - it's a single, connectable point of control for a value that otherwise scatters across your graph.

The inputs and outputs that matter

There's really just the seed value you set, and the seed number it outputs. You route that output to anything that consumes a seed. That's the node.

The reason it exists at all is connectivity - native KSamplers keep their seed as an internal widget you can't easily share, so a standalone Seed node gives you a wire to fan out from.

How to install it

Ships with WAS Node Suite. Install the pack once - ComfyUI Manager, search WAS Node Suite, install, restart - or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui
pip install -r was-node-suite-comfyui/requirements.txt

then restart. On Windows portable, run the pip step with python_embeded\python.exe -s -m pip install -r ... (or install.bat). No models, no dependencies of its own - it's a number holder.

Common issues & troubleshooting

Your images stopped varying. If you wired a fixed Seed into every sampler and set the graph to not randomize, every run produces the same image - by design. That's great for reproducing a result and annoying when you actually want variety. To explore, either bump the seed each run or feed a Random Number node into your seed source instead.

Randomize behavior differs from native nodes. ComfyUI's built-in seed widgets have that familiar control-after-generate menu (fixed / increment / randomize). A standalone Seed node behaves like whatever you wire into it, so if you want auto-randomization, drive it from a random source rather than expecting the native dropdown here.

Modern graphs often use rgthree's Seed instead. The rgthree pack ships a popular Seed node that remembers the last value it ran, which some people prefer for iterating. WAS's Seed is perfectly fine, but if you already run rgthree for its other quality-of-life nodes, you may not need this one - pick whichever your graph already leans on.

The suite won't load. WAS Node Suite is a large pack, unmaintained since late 2023, and the classic symptom is an "Import Failed" after a ComfyUI update, from a dependency version clash (opencv especially). Reinstall the requirements against your ComfyUI Python and restart. The Seed node is trivial; the suite install is the only thing that tends to break.