Random Seed

This is the plainest node in the whole pack, and that's the point. RiceRoundRandomSeedNode takes zero inputs - no fields to fill in, nothing to configure - and just outputs a fresh INT every time the graph runs. Wire it straight into your KSampler's seed input and every execution of your published workflow gets a genuinely new seed, without you or an end user having to think about it.

Why a seed node with no controls

A seed is what makes diffusion sampling reproducible: same seed, same prompt, same settings, same image - it's the value everything else in the noise-generation process is deterministic against. Most tools give you a widget for it because sometimes you want to lock a seed and change one other variable to see what shifts (the standard "seed farming" workflow the community uses for comparing samplers and settings). RiceRound's version skips all of that. There's no control_after_generate dropdown, no min/max, no way to pin a specific value from this node - it exists purely to hand a random INT to whatever needs one.

That fits the context it's built for: a workflow published to riceround.online is meant to be run repeatedly by whoever's using the generated page, and you almost certainly don't want every one of those runs reusing the exact same seed you happened to leave in the graph when you published it. This node is the "don't make me think about it" answer to that - every run gets its own seed automatically, and if you want a specific seed to be user-adjustable instead, that's what RiceRoundIntNode is for, wired into your sampler the same way.

The output that matters

There's exactly one: an unlabeled INT, ready to plug into a KSampler's seed field or anywhere else in your graph that expects an integer seed. That's the entire interface.

Installing it

Through ComfyUI Manager: search "RiceRound," install, restart. Or clone it directly:

cd ComfyUI/custom_nodes
git clone https://github.com/RiceRound/ComfyUI_RiceRound

then restart ComfyUI. The README doesn't mention any dependencies or model downloads for the pack as a whole - it's a workflow-packaging and publishing layer, not a generation pack, so this install should be quick regardless of which of its nodes you're actually here for.

Common issues & troubleshooting

You want to reproduce a specific result and this node won't let you pin a seed. That's by design, not a missing feature - this node is deliberately unconfigurable. If reproducibility matters for a given run, either note the seed some other way (it's not surfaced anywhere by this node itself) or swap it out for RiceRoundIntNode set to a fixed value, which does give you a controllable field.

Is it safe to leave in a workflow you're about to publish? Yes - it's an Input-category node, and per the README that's exactly what belongs in the graph you run through RiceRound Publish. The nodes to strip before publishing are the separate Output-category debugging nodes, not this one.

Every test run during development produces a different image and it's hard to compare changes. That's this node doing exactly what it's built to do - randomizing every execution. While you're iterating on the rest of the workflow and specifically want a fixed comparison point, temporarily swap in a plain INT source (or RiceRoundIntNode) instead, then switch back to this once you're ready to publish and want genuine per-run randomness again.