LoRA Stack With Triggers

The LoRA stack is one of the genuinely useful patterns in ComfyUI - the idea rgthree's Power Lora Loader made famous, which is exactly who AUN credits for inspiration. Instead of one LoRA per node, you get a single node with slots, and it applies them all to your model at once. AUNLoraStackWithTriggers is that pattern with a twist: every slot carries its own trigger words, and the node stitches them together so your prompt can be auto-populated with exactly the trigger words your active LoRAs need. The catch is in the name - this node is deprecated, and the pack wants you on its successor.

What it does

Up to 10 slots, each with a lora_N, strength_model_N, enabled_N, and trigger_N. Set num_slots to show only the ones you need. trigger_joiner is the string used to join triggers from active slots (default a comma), and dedupe_triggers removes repeats across slots and your base_prompt while preserving order - a nice touch, because the same trigger word on three LoRAs is a real occurrence.

The outputs are what make it more than a stack: MODEL (all active LoRAs applied), selected LoRAs (which ones are active, as a string), labels, trigger words, and trigger + prompt - the trigger words joined onto your base_prompt, ready to feed a CLIP Text Encode. That last one is the feature: set your base prompt once, and the node keeps the trigger words in sync with whichever LoRAs are enabled.

The apply_stack toggle is worth knowing: off, it returns the input model unchanged and suppresses the trigger text outputs - a clean way to A/B "LoRAs on vs off" in one switch.

Why it's deprecated

The successor, AUNLoraStackWithTriggersModelClip, adds per-slot CLIP strength (this node is model-only, so a LoRA that needs a separate CLIP-side strength - and plenty do - is only half-applied), plus compact mode with an overlay UI and drag-to-swap. The pack says this node will be removed in a future release. Keep it for old workflows; for anything new, use the Model Clip version.

Installing it

Ships in the AUN pack:

• ComfyUI Manager: search "AUN ComfyUI Nodes", install, restart.
• Manual: cd custom_nodes && git clone https://github.com/loz2754/AUN-ComfyUI-Nodes, then restart. Manual installs and ModuleNotFoundError: cv2? pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt.

Common issues

The model-only limitation is the one that actually bites: if your stacked LoRAs include style or character LoRAs that affect text encoding, you'll get images that look half-applied because CLIP never got patched. That's the reason to move to the successor. Otherwise, watch the trigger output - with apply_stack off it goes silent, which confuses people who expected the triggers regardless. If your prompt suddenly loses its trigger words, check that toggle first.