Seed Control

Stock ComfyUI treats your seed as a number in a box. That's fine until you've got five of them and can't remember which number produced which image. Seed Control, from the A1rCHAN A1RWorkshop pack, is a seed node that remembers its recent history and shows you the actual image each seed made, then queues the run for you when you pick one. It's the difference between "I had a good seed somewhere" and pointing at it.

How it works

The Python side is a pure pass-through: one seed INT input (0 to 2^64−1, default 0), one seed INT output. All the personality lives in the frontend, where the node grows two buttons on its face:

• manual random - rolls a fresh random seed into the box. No control_after_generate here, so "randomize" stays a deliberate act instead of a hidden dropdown that fires after every run. If you've ever lost a good seed to the after-generate trap, that's a feature.
• pull history - opens the node's recent-seed list. The node keeps up to 10 seeds it has seen, and each entry gets a popover preview of the image it generated, captured from the queue. Click one and it restores that seed and queues the prompt.

Per the README: "These two buttons will queue prompt when clicked." So this is a "look, choose, generate" node, not just a storage box - clicking either button fires a run with that seed.

The history is kept per-node and serialized into your workflow, so a saved workflow drags its seed history along with it.

The inputs and outputs that matter

Honestly, just one input and one output:

• seed (INT) - the current value. You can type into it, or let the buttons manage it.
• Output seed (INT) - wire this into the seed input of your sampler. Because it's an output socket, one Seed Control can fan out to several samplers, giving you "one seed, five samplers" the way a typed-in KSampler widget can't.

If your KSampler's seed is still a widget, right-click it and convert it to an input first, then drop this node's output onto the socket.

Installing it

This is part of the A1rCHAN/ComfyUI_A1rSpace pack (A1RWorkshop). It isn't on the official ComfyUI Registry yet, so Manager search may miss it - the reliable path is a manual clone:

cd ComfyUI/custom_nodes
git clone https://github.com/A1rCHAN/ComfyUI_A1rSpace

Restart ComfyUI and it appears under A1R Workshop → Config Pads. No Python dependencies (requirements.txt is empty) and no models to download. The real requirement is a recent ComfyUI: the pack is built on the new V3 node API (comfy_api.latest) and the Nodes 2.0 frontend. Old installs will fail to import the whole pack.

Common issues

The honest caveat: this pack is a one-person project with essentially no community footprint - a Reddit corpus search for the pack name returns nothing. So there's no collective troubleshooting to lean on; the practical risks are the frontend ones. If the history popover's image previews look wrong or missing, that's a rendering issue on your UI version - the pack targets Nodes 2.0 and only promises basic compatibility with the legacy canvas. And remember the buttons queue a prompt: clicking "manual random" fires a generation, so it's not a hands-off-until-you're-ready button. For the "remember what I ran" half of it, it's genuinely handy; for the "don't touch my seed" crowd, the deliberate-randomize design is actually a selling point.