Lora Stack Combiner (LoraManager)

This is a glue node, and a genuinely handy one once you start splitting your LoRA selection across several sources. It takes two LORA_STACK inputs and welds them into a single stack. That's it. But that one operation is what lets you keep your "always-on" style LoRAs in one Stacker and your "this run only" character LoRAs in another, then combine them right before they hit a loader.

The reason you need a dedicated node for it: a LORA_STACK is a structured object, not a string you can concatenate. You can't just merge two of them with a text node. And most loaders take exactly one stack input, so if you've built two - say, one from a Lora Stacker and one from a Lora Randomizer - you need something to join them first. That's this.

A concrete pattern where it shines: keep a fixed "house style" stack that goes into every render, and a separate variable stack from a Randomizer or Cycler that changes run to run. Combine them right before the loader and you get consistent styling with rotating content, without duplicating the fixed LoRAs into every branch.

How it works

Feed it stack A and stack B. It appends B's entries onto A's and outputs the combined LORA_STACK, preserving each LoRA's strength. Order matters a little: LoRAs generally apply in sequence, and the troubleshooting lore is that when stacking triggers black images or NaNs, putting the larger LoRAs first can help - so if you're chaining combiners, keep an eye on which stack lands first.

Inputs and outputs that matter

• lora_stack_a (LORA_STACK, required) - the first stack.
• lora_stack_b (LORA_STACK, required) - the second; its entries get appended after A's.

One output: LORA_STACK - the merged result. Send it to any loader with a stack input, or into another combiner if you're joining three or more sources.

How to install it

Via ComfyUI Manager: search lora-manager, Install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/willmiao/ComfyUI-Lora-Manager.git
cd ComfyUI-Lora-Manager
pip install -r requirements.txt

then restart. Nothing heavy - it's a management pack, so the install is light Python.

Common issues & troubleshooting

Both inputs are required. If you only have one stack, you don't need this node - wire that stack straight into your loader. The combiner is for when you genuinely have two.

Duplicate LoRAs. If the same LoRA appears in both stacks, you'll likely end up with it applied twice, stacking its strength. That's usually not what you want - dedupe upstream or keep each LoRA in exactly one source stack.

Nothing visibly changes after combining. The combined stack is still just a description until a loader applies it. And remember the usual gotcha: LoRAs only fire on their matching base architecture, so a stack that mixes families will quietly drop the mismatched ones at load time.