Multi LoRA Combine

Multi LoRA Combine is the mirror image of Multi LoRA Splitter: where the splitter turns one MULTI_LORA_STACK into four plain stacks, this node takes plain stacks and folds them back into a single multi-slot payload. It's the node you reach for when your graph produces LoRA stacks in several places - a saved stack from Prompt Manager here, a hand-built stack there - and you want them all riding in one MULTI_LORA_STACK before it hits a recipe or another stacker.

How it works

Every input is optional. It starts from an optional multi_lora_stack payload and appends the four optional LORA_STACK inputs - lora_stack_a, b, c, d - to their matching slots. Slot A from the incoming payload stays in slot A, and any connected lora_stack_a gets appended after it. Slots you don't touch come through unchanged. The single output, multi_lora_stack, is a dict-shaped payload with a/b/c/d keys, ready for anything that consumes that type.

The useful property here is that it's a merge, not a replace. An existing payload's slot A keeps its LoRAs and your new stack's entries get appended to the same slot, so you can build a multi-slot configuration incrementally: seed it from a saved prompt's stack, then add this run's style LoRA on top without touching the saved entry. Order within a slot is preserved, and since LoRA application order matters (the first entry in a stack applies first), that's not just a nicety.

Wiring it

The typical pattern:

1. Connect a MULTI_LORA_STACK source (Prompt Manager's multi_lora_stack output, or a Multi LoRA StackerLM) to the multi_lora_stack input.
2. Connect loose LORA_STACK outputs - from LoraStackCombine, a saved stack, or another loader - to the matching a/b/c/d inputs.
3. Route the output into a Prompt Manager node, a Recipe Builder's multi_lora_stack input, or through Multi LoRA Splitter when you need plain stacks again.

Installing

Part of the ComfyUI-Prompt-Manager pack, so install it once for all of these nodes:

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

Or use ComfyUI Manager (search "Prompt Manager"), then restart ComfyUI.

Gotchas

Two things trip people up. First, the inputs are typed LORA_STACK - connect a MULTI_LORA_STACK where a plain stack is expected and ComfyUI will complain about the type mismatch, so put a splitter in between if you're going the other direction. Second, because everything is optional and merging is additive, it's easy to accidentally double-apply a LoRA: if a stack is already inside the incoming payload's slot and you also connect it to that slot's input, you get it twice. When a generation suddenly looks overbaked and you can't figure out why, check for that.