Multi LoRA Splitter

Multi LoRA Splitter is the unglamorous half of the pack's multi-slot LoRA system, and it exists because the glamorous half needs it. The pack's stackers produce a MULTI_LORA_STACK - a payload that bundles four named LoRA stacks under the slots a/b/c/d. That payload is compact and travels well through Prompt Manager and the recipe nodes, but the rest of ComfyUI doesn't speak it. Standard nodes want plain LORA_STACK wires. This node is the adapter: one input, four outputs, zero decisions.

How it works

It takes a single multi_lora_stack input and splits it into four LORA_STACK outputs: lora_stack_a, lora_stack_b, lora_stack_c, lora_stack_d. Empty slots come out as empty stacks rather than errors, so you can feed a payload that only fills slots A and B into this node and the C and D outputs are simply empty lists - safe to wire into a loader that expects a stack, since an empty stack just means "no LoRAs here."

The mechanism is boring in the best way: it reads the a/b/c/d keys out of the dict-shaped payload and normalizes each into a standard list of (path, model_strength, clip_strength) tuples. If you hand it something that isn't a proper multi-stack at all, it returns four empty stacks instead of crashing. That's the entire node.

Why you'd actually use it

Typical flow: Multi LoRA Stack builds one multi-slot payload for your whole generation, and you want slot A on the image model while slots C and D feed the video or text-encoder path. Run the payload through this splitter and each branch of your graph gets its own stack without you re-adding LoRAs in three places. It's also the natural companion to Multi LoRA Combine, which does the reverse (merging loose stacks back into one payload) - the pair makes the slots "real" anywhere in your workflow.

Wiring and install

Wire the multi_lora_stack output from Multi LoRA StackerLM (or any node producing that type, including Prompt Manager's multi_lora_stack output) into the input, then route each output to the lora_stack input on a standard LoRA loader. Install via ComfyUI Manager (search "Prompt Manager") or:

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

Restart ComfyUI after install. No models, no dependencies beyond the pack itself.

One honest caveat

This is a utility node, not a feature. If you're not already using multi-slot stacks - which mostly pays off for Wan/multi-model pipelines and saved multi-stack prompts - you don't need it, and you shouldn't add it for the sake of tidiness. Where it earns its keep is in workflows that swap saved multi-stack prompts in and out, because the splitter is what lets a single saved payload drive several different graphs. If that's your situation, it's the boring node you'll reach for constantly.