LoRA Stack Combine

Sometimes you don't need the whole multi-slot apparatus - you just need more than one LoRA stack wired into a single LORA_STACK input. That's LoRA Stack Combine: up to four LORA_STACK inputs, one concatenated LORA_STACK output. If your ComfyUI career has hit the "why is there only one lora_stack socket on this loader" wall, this is the three-line fix.

How it works

All four inputs - lora_stack_a, b, c, d - are optional. The node appends them in order, a→b→c→d, into one output stack and normalizes each entry into the standard (lora_path, model_strength, clip_strength) tuple form as it goes. Entries from stack A come first, then B, and so on, so the ordering of your output reflects the order you connected the inputs.

That ordering matters more than it looks like it does. LoRA stacks apply sequentially - the first entry in the stack is applied first - and stacking order changes the result, especially when two LoRAs touch the same concepts. This node preserves connect-order, so if you want your character LoRA applied before your style LoRA, connect character to A and style to B, and don't shuffle the wires later and expect the same output.

Why it exists

Mostly workflow tidiness. A Wan or multi-pass graph can end up with LoRA stacks built in different places - one from a saved prompt entry, one hand-assembled for the current run - and a single loader only takes one. This merges them at the end without you editing either source. It's also the plain-stack equivalent of Multi LoRA Combine: use it when you're working with ordinary LORA_STACK wires and don't need the a/b/c/d payload format at all.

Wiring and install

Connect the stacks to the inputs, run the lora_stack output into any LORA_STACK input (a standard LoRA loader, or a Prompt Manager node's lora_stack_a). Empty inputs contribute nothing, so a partially wired graph just works.

It ships in the ComfyUI-Prompt-Manager pack:

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

Or install via ComfyUI Manager (search "Prompt Manager"), then restart. No models or extra dependencies.

Gotchas

The honest caveat: this is a pure concatenation node, not a deduplicator. If the same LoRA shows up in two inputs - say, saved in the prompt entry and added to the current run - it gets applied twice, with all the strength-stacking that implies. Also worth knowing: LORA_STACK and MULTI_LORA_STACK are different types here, so don't expect to merge a multi-slot payload with this node; that's Multi LoRA Combine's job. Keep it for the simple case, and it'll never bite you.