Quick Lora Stack (x3)

If Lora Stack (x3) is the full-featured version, Quick Lora Stack (x3) is the version that admits most LoRAs never need a separate clip weight. Same three slots, same enabled toggles, same chaining - but each slot drops the clip_weight dial entirely and just has model_weight. The one output stack is still a first-class lora_stack, so everything downstream treats it identically.

The mechanism is worth knowing because it explains the name. When a LoRA is added, the quick version writes your model_weight into both the model and clip slots of the stack entry. So model_weight 0.8 becomes (path, 0.8, 0.8). You're not applying the LoRA to the clip more weakly or strongly - you're just saying "same everywhere," which is what most LoRAs want anyway. The full Triple Lora Stack node lets you split those dials; this one assumes you don't care to.

That's the whole decision, really. Reach for this one when:

• You're stacking several LoRAs and want the smallest possible node face, not six weight fields per stack.
• You're on an architecture where clip weight is a non-issue anyway. The KB's training lore notes the modern shift: text-encoder training is largely gone from current recipes, and for a lot of recent models the clip side of a LoRA is either irrelevant or handled identically.
• You want to build stacks fast and tune later.

Skip it and use the full version if you're one of the people who deliberately runs SDXL character LoRAs with a weakened clip weight, or you know a specific LoRA in your stack specifies different model/clip values in its Civitai metadata. Then the extra dial earns its pixels.

Everything else is the same family behavior: enabled toggles skip an entry instead of forcing you to delete it, the optional lora_stack input chains three-slot nodes into bigger stacks, and the output feeds Sage Utils' Lora Stack Loader to patch a model and/or Construct Metadata so your LoRA choices land in saved image metadata. The pack also ships x6 and x9 quick variants if three slots are too few.

Install is the pack standard - ComfyUI Manager (search "Sage Utils") or:

cd ComfyUI/custom_nodes
git clone https://github.com/arcum42/ComfyUI_SageUtils.git

then restart. No downloads, no heavy deps beyond the pack's dynamicprompts. LoRA dropdowns fill from your models/loras folder, so a fresh clone shows whatever's already in there.