[SK] Power LoRA Loader

If you've used rgthree-comfy, that name will ring a bell - rgthree's own Power Lora Loader is probably the single most-borrowed node in the entire ComfyUI custom-node ecosystem, the thing people install rgthree's whole pack for even when they don't want anything else in it. This is a different implementation of the same idea from a different, much smaller author: one node, five LoRA slots, each with its own enable toggle and independent model/CLIP strengths, so you stack a set of LoRAs without chaining five separate loader nodes across your canvas.

The honest comparison: rgthree's version is an unlimited, dynamically-expandable stack with a right-click info dialog that fetches trigger words straight off CivitAI. This one is fixed at five slots, each a plain widget rather than a dynamic UI element, and it's part of SK-ComfyUI-FolderingLoader - a small, undocumented pack whose whole differentiator is a folder-tree file picker on every combo, including each of these five LoRA slots. If you're already running rgthree-comfy, you probably don't need this. If you're not, and five is enough, this gets you most of the convenience in a lighter package with browsable folders on top - a genuinely reasonable trade for a simpler pipeline.

How it works

Same underlying mechanism as any LoRA loader, run five times in sequence inside one node: each enabled slot reads its LoRA file and merges its weight adjustments into the model and CLIP you passed in, at that slot's own strengths, before handing the result to the next slot. Disabled slots are skipped entirely - flip lora_N_enabled off and that slot's file, even if one's selected, never gets applied. That's what makes it useful for A/B testing a stack: leave all five wired up and just toggle which ones are live.

The inputs that matter

• model / clip - the pair every slot modifies in sequence.
• lora_1 through lora_5 - each a folder-tree combo, independent of the others.
• lora_N_enabled - a checkbox per slot, default off. Nothing applies until you flip it on, even if a file's already selected in the combo next to it.
• lora_N_strength_model / lora_N_strength_clip - per-slot strengths, default 1.0 each, same -100 to 100 range as the single-LoRA loaders. Independent per slot, so slot 1 can run hot while slot 3 runs subtle.

Outputs are the fully-stacked MODEL and CLIP, ready for your sampler and text encode as usual.

Installing it

No install section shipped with the pack, so it's the standard two routes: ComfyUI Manager → search SK-ComfyUI-FolderingLoader or the class name → install → restart, or manually cd ComfyUI/custom_nodes && git clone https://github.com/SKFRMSEHF/comfyui_SK_Loader and restart. No bundled models - bring your own LoRA files.

Common issues

• Forgetting to flip the enable toggle. The single most likely "why isn't my LoRA doing anything" moment with this node - a file selected in the combo does nothing until lora_N_enabled is checked.
• Only five slots, no more. If your stack needs a sixth LoRA, chain a second SK_PowerLoraLoader (or a plain SK_LoraLoader) after this one rather than looking for a way to add slots - there isn't one, unlike rgthree's dynamically-expandable version.
• Stacking gets unpredictable fast. Multiple LoRAs interacting is inherently less controllable than one at a time - if output goes strange with several slots enabled, disable them one at a time to isolate which one's causing it, rather than guessing at strengths.
• Base model compatibility still applies per slot. Each slot's LoRA still has to match your checkpoint's architecture; mixing an SDXL LoRA into a Flux stack in one of the slots won't work just because the node makes it easy to select.
• No community trail. This pack has essentially no online footprint to search when something's unclear - the file itself, and rgthree's much better-documented equivalent, are your best reference points for what "normal" LoRA-stacking behavior looks like.