Seed History

You know the sequence: you batch out twenty seeds, one of them is the one, and then you spend ten minutes trying to remember which seed produced it because you already rolled the widget past it. Seed History is the fix. It's a seed node with a memory - every seed you run gets logged with a timestamp, and you can click any entry in the history to reload it instantly.

How it works

Underneath it's a plain seed node: you give it an INT, it hands you back a validated INT for your KSampler. The history is a JavaScript layer on top, courtesy of the pack's seed_history_ui.js frontend. Each run records the seed with a human-readable timestamp ("5m ago", "2h ago"), keeps the list capped at ten entries, and dedupes rapid repeats within a 500ms window so spamming Generate doesn't flood the list.

The persistence is the part that surprised me: history lives in the browser's localStorage, not in the workflow file. That means it survives ComfyUI restarts and survives you closing the browser tab - the seeds you tried yesterday are still there this morning. On the backend the node validates the value (it'll fall back to 12345 rather than choke on garbage) and passes it through to the seed output. It supports the standard control_after_generate behavior too, so you can keep incrementing or randomizing after each run.

The one input

• seed - INT, 0 to 4,294,967,295, default 12345. That's the whole interface. The history, the Generate button, the Clear button, the click-to-restore - all of it is frontend chrome around this single value.

Output: seed, which wires straight into your KSampler's seed input.

Why you'd reach for it

Seed farming is the use case. The community pattern is to run a sampler at low steps (DPM++ 2M or UniPC, 10–15 steps) across a pile of seeds to find compositions worth committing to, then re-run the winners at full quality. That workflow is a nightmare with a stock seed widget because you lose track of which seed was which. With Seed History, the good ones are one click away - no copying numbers into a text file.

Installing it

It's part of the ComfyUI-KikoTools pack. In ComfyUI Manager, search for "ComfyUI-KikoTools" and install; restart ComfyUI and look under ComfyAssets → Seeds. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/ComfyAssets/ComfyUI-KikoTools.git
cd ComfyUI-KikoTools
pip install -r requirements.txt

Restart and the node appears. No extra model files, no API keys.

Gotchas

Two things trip people up. First, the history is per-browser: if you switch from Chrome to Firefox or clear your browser storage, your history is gone - the seeds themselves don't get baked into the workflow JSON. Second, if you click a historical seed and the output looks identical to your last run, check your KSampler's other inputs first - same seed only reproduces an image when the sampler, scheduler, steps, CFG, and model are all identical too. The node remembers the seed, not the whole recipe.