Quick Lora Stack (x6)

A serious Pony or anime workflow routinely stacks five or six LoRAs: character, style, quality, a couple of merges. Chain six individual LoRA loaders and your graph is a vertical wall of nodes, each with its own weight widgets and no easy way to flip one off for a comparison run.

Quick Lora Stack (x6) crams six LoRAs into one node - each with its own name dropdown, its own weight, and its own on/off switch - and emits a single lora_stack you can chain further or feed straight to a loader.

How it works

The node is one slice of the Sage LoRA-stack family. The x6 variant gives you six identical triples of inputs:

• enabled_1 … enabled_6 - a boolean per slot, all defaulting to on.
• lora_1_name … lora_6_name - the LoRA file dropdowns.
• model_1_weight … model_6_weight - one weight per LoRA, default 1.0, range -100 to 100.

The optional lora_stack input appends onto an existing stack, so you can put an x6 node after an x3 node and build a nine-LoRA stack out of readable blocks instead of spaghetti. The output lora_stack is the same LORA_STACK type the whole family speaks.

Quick here means the same thing it does everywhere in this family: each entry writes one weight to both the model and clip slots, rather than exposing a separate clip weight per LoRA. Flip enabled_2 off and that LoRA simply isn't added - the rest of the stack is unaffected.

Where it fits

The x6 hits the sweet spot for a lot of users - it's exactly the "base stack" size for character + style + detail + refinement setups. Wire its output into Load Models + Loras or Lora Stack Loader to apply everything at once, and into Construct Metadata to record the whole stack in the PNG's metadata so your LoRA list travels with the image.

The on/off switches are the killer feature. They're the reason this node family exists - the author built them after a Reddit request for LoRA loaders with boolean switches. When you're A/B-testing whether a style LoRA actually helps, you flip a boolean and queue, not rewire the graph.

Installation

Sage Utils, same pack install as the rest:

cd ComfyUI/custom_nodes
git clone https://github.com/arcum42/ComfyUI_SageUtils
cd ComfyUI_SageUtils
pip install -r requirements.txt

Or via ComfyUI Manager, search "Sage Utils", then restart ComfyUI. Only dependency is dynamicprompts.

The honest take

Six slots is plenty for most workflows, but if you're one of the "I could use nine" people, the x9 variant exists with the exact same pattern. And if a LoRA genuinely needs a clip weight that differs from its model weight, the quick family can't express that - reach for the full Lora Stack (x3/x6/x9) nodes, which expose both weights. For everyone else, one weight per LoRA is the right amount of decision, and the booleans make experimentation cheap.