Apply LoRA+

Apply LoRA+ is the node you reach for when your workflow has more LoRAs than wires. Where the built-in LoraLoader applies exactly one LoRA per node, this one takes a whole stack - a list of (name, strength) tuples - and applies them all to your model in one shot, with one global strength knob on top. If you've ever stared at a wall of four LoraLoaders chained together and wished it were a single node, this is it.

It's part of ComfyUI-FBnodes, François Beaudry's grab bag that also includes Save Video+ and the Load Model+ family. Same author as ComfyUI-Prompt-Manager, same honest, utility-first vibe: nothing here generates pixels, it just makes the graph you already have less annoying to build and read.

How it works

The lora_stack input is deliberately loose, so you can feed it three different things and it'll cope with all of them:

• a LORA_STACK - the list of (path_or_name, strength) tuples that Prompt Manager Advanced and the Visual LoRA Picker+ output
• a MULTI_LORA_STACK - the {a,b,c,d} payload some stacks ship, which it flattens in slot order
• a plain multiline STRING, one LoRA name or full path per line, each applied at strength 1.0

Under the hood it resolves each entry against your LoRA folder, applies the LoRAs in order through ComfyUI's normal LoRA mechanism, and skips anything it can't find on disk rather than crashing. The strength input (0–10, default 1.0) multiplies every entry's model strength, so you can build a stack once and then dial the whole thing down for a weaker overall effect without touching each entry. There's no CLIP input here - this one only touches the MODEL, so it's the right choice when you're stacking style LoRAs and don't need the text-encoder half.

The two inputs that matter

• lora_stack (*) - where your stack comes in. Wire it from a stack-building node or paste paths.
• strength (FLOAT, 0–10, default 1.0) - the global multiplier. 0 disables everything; useful for A/B testing a stack against no LoRAs at all.

Output is a single model (MODEL) that feeds straight into your KSampler or, for LTX video work, the LTX-specific sibling Apply LTX LoRA+ handles the audio/video split instead.

Installing it

ComfyUI Manager is the easy route - search "ComfyUI-FBnodes" and install. Manual:

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

Then restart ComfyUI. The pack's real dependency is av (PyAV) for video encoding plus color-matcher; both are in requirements.txt. No model downloads - this is a pure plumbing node.

Gotchas

Where people get burned: if your stack entries are names rather than full paths, the fuzzy resolver needs the LoRA to be somewhere ComfyUI can find it under models/loras/. And if you chain two Apply LoRA+ nodes with overlapping stacks, remember LoRA application isn't idempotent - stacking the same LoRA twice is stronger, not the same. Want a quick visual picker instead of filenames? The pack's Visual LoRA Picker+ builds the stack for you; this node is what consumes it.