LoRA Stacker

Stacking multiple LoRAs the vanilla way means chaining a Load LoRA node for each one - model in, model out, model in, model out, a little conga line across your canvas. LoRA Stacker collapses that into a single node where you just list the LoRAs and their weights. It doesn't apply them itself; it builds a tidy LORA_STACK bundle that the Efficient Loader (or another stack-aware node) applies in one shot.

That's the pattern the whole pack is built on - describe the thing in one node, apply it downstream - and for LoRAs it's genuinely nicer than the wire spaghetti.

How it works

You set lora_count to however many LoRAs you want, and the node shows that many rows. Each row picks a LoRA and its strength. The assembled list comes out the LORA_STACK output, which you plug into the lora_stack input on an Efficient Loader / Eff. Loader SDXL. You can also feed another stack into this node's optional lora_stack input to extend an existing stack rather than start fresh - that's how you build up a big stack across nodes.

The inputs that matter

• input_mode - simple or advanced. In simple, each LoRA has one weight (lora_wt). In advanced, you split it into separate model_str and clip_str dials, so you can, say, run a strong model effect with a lighter CLIP effect. Most people live in simple mode; advanced is there when you need that separation.
• lora_count - how many rows appear. Bump it up as you add LoRAs.
• lora_name_1, lora_wt_1, … - the per-row picker and weight. In advanced mode you also get model_str_N and clip_str_N per row.

The only output is LORA_STACK. It goes into a loader's lora_stack input - that's the entire job.

Installing it

ComfyUI Manager → search Efficiency Nodes for ComfyUI, or:

cd ComfyUI/custom_nodes
git clone https://github.com/jags111/efficiency-nodes-comfyui

Restart after. Your LoRA files need to be in ComfyUI's usual models/loras folder to show up in the dropdowns.

Common issues

The most frequent "why is nothing happening": a LoRA Stacker on its own does nothing visible - it only builds a list. If you don't wire its LORA_STACK into a loader that actually applies the stack, your LoRAs never touch the model. Check that connection first.

Second, a real and long-documented quirk of this pack: on the SDXL base+refiner path, LoRAs from the stack have historically only affected the base model, not the refiner. If a stacked LoRA looks weaker than it should on an SDXL base+refiner run, that's the likely reason - it's a known behavior, not your settings.

And the pack-level stuff: if the node doesn't show up, look for IMPORT FAILED in the console (often a pip freeze error) and update ComfyUI, the packages, and the node. This pack breaks on ComfyUI frontend updates fairly often; updating it to the latest commit is the standard fix.