Fooocus LoraStack

Every ComfyUI power user ends up with a favourite multi-LoRA loader. rgthree's Power Lora Loader is the community darling, but if you're inside this pack you might as well use the one that's built for it: Fooocus LoraStack stacks up to ten LoRAs with per-slot strength and hands them to the Fooocus Loader in a single wire. Same idea, no extra pack.

How it works

The node's real trick is the num_loras field. It's a configurator: set it to 5 and the node spawns five lora_x_name + lora_x_strength pairs; set it to 3 and the rest drop away. That's the mechanism that makes a giant node usable - you never wade through ten empty slots, you only see the ones you asked for. Each named slot pulls from your LoRA folder, and each strength runs -10 to +10 (negative strengths are a legit trick for suppressing a trained-in style).

The inputs that matter:

• toggle - on/off for the whole stack. Off passes the optional stack through untouched, which is your "disable all LoRAs" switch without editing anything.
• num_loras - 0 to 10. How many slots to show. This is the dial you'll actually touch.
• lora_x_name / lora_x_strength - the per-LoRA pair, x being the slot number.

Optional input: optional_lora_stack - lets you chain another stack in, so you can build a stack across multiple nodes and merge them.

Output: lora_stack (type LORA_STACK) - and the only place that plugs in is the Loader's optional_lora_stack input. Wire it there, not into a standard model loader.

Installing

Pack install: ComfyUI Manager, search "Fooocus Nodes", or clone the repo into custom_nodes. No model downloads - the LoRAs themselves come from your own ComfyUI/models/loras folder, same as anywhere else.

The honest comparison

The Fooocus LoraStack is genuinely fine, but let's not oversell it: rgthree's Power Lora Loader wins on the extras - trigger words, CivitAI lookup, per-entry toggles. What this stack has over it is that it's already here, it's shaped for the pack's pipe, and it's dead simple. If you're running a full Fooocus-pipe workflow, it's the natural choice. If you're building general-purpose ComfyUI graphs, don't install this pack for the LoraStack - get rgthree and save yourself a pack's worth of baggage. Stacked LoRAs are load order-sensitive in a way you'll only notice at high strengths; keep total weight reasonable (2-3 active LoRAs at moderate strength) and the stack behaves.