RedNode LoRA Stack Save

RedNode LoRA Stack lets you build a multi-LoRA rig with per-slot strengths, random ranges, and trigger words. This node is the other half of that workflow: it takes the stack you've built and writes it to disk under a name, so it shows up as a preset you can reload from the stack's preset dropdown in any workflow. It's the persistence layer for the whole arrangement.

Two inputs, and the first one has a trick to it. name is the preset name you're saving under - saving over an existing name overwrites it, which is how you update a stack you've improved. stack_json is the slot list, and here's the detail the tooltip is pointing at: wire the stack node's own stack_json here by converting that widget to an input. That way you save exactly what the stack node is currently holding - including the toggles, random ranges, and keywords - rather than a hand-copied snapshot that drifts from the real thing. It's the same "Convert widget to input" mechanic that solves every stale-duplicate problem in ComfyUI: one authoritative source, wired through.

Now the part that will bite you exactly once. This is an output node, which means it runs every time you queue the workflow. Every generation would re-save the preset - overwriting it with whatever the stack happens to hold at that moment, which is almost never what you want. The README and the node description both say the same thing in caps: keep this node muted except when you actually want to save. Build your stack, mute the save node, and forget it. When you've dialed in a stack worth keeping, unmute, queue once, mute again. That muted-state workflow is the design, not a workaround.

Why bother saving at all when you could just copy the stack_json into a note? Presets are the reusable form: they show up in the stack node's dropdown (after a node-definition refresh, r), they survive across workflows in the same ComfyUI install, and they travel with the pack's user-data directory. If you find yourself rebuilding the same character-plus-style-plus-vae stack in every new workflow, one save ends that ritual permanently.

It's a thin node - a save button with a name field - and it doesn't need more than that. The word count here is on purpose: there isn't much to it, and the entire skill is remembering the mute toggle and wiring stack_json from the source rather than retyping it.

Installation is the pack:

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

or search RedNode Studio in ComfyUI Manager, restart, done. No pip dependencies; works with any model.