LoraLoaderStackedVanilla

If you want to run multiple LoRAs at once and also get their trigger words out as data, LoraLoaderStackedVanilla is the no-frills version of the pack's stacked loader. It does everything LoraLoaderStackedAdvanced does - no MODEL input, builds a LORA_STACK you chain forward, emits civitai_tags_list and meta_tags_list per LoRA - minus the dropdown preview and the right-click info dialog. In exchange it has zero dependency on pythongosssss/ComfyUI-Custom-Scripts, which makes it the version to use if you're trying to keep your install lean or if Custom-Scripts' features don't do anything for you.

The "Vanilla" label in this pack is honest about what you're giving up: only the front-end. Same stacked-loading mechanics, same tag lookups, same LORA_STACK output type that's compatible with the efficiency-nodes family of loaders this pack explicitly targets.

How it works

The inputs mirror the Advanced version exactly: lora_name (dropdown of your loras folder), lora_weight (default 1.0), force_fetch, append_loraname_if_empty, plus optional lora_stack and override_lora_name. Wire an existing stack into lora_stack to keep chaining, and it appends this LoRA to the incoming list. The three optional items are where the flexibility lives: no stack connected and it starts one; override_lora_name fed from a LoraListNames node and it ignores the dropdown.

Four outputs. civitai_tags_list and meta_tags_list are the LIST outputs feeding TagsFormater/TagsSelector - Civitai trainedWords by hash (cached in loras_tags.json), and safetensors training tags sorted by frequency, respectively. LORA_STACK is the accumulated stack that carries forward to whatever actually applies the LoRAs. lora_name is the resolved filename.

The two booleans behave like everywhere else in this pack: force_fetch re-queries Civitai even when cached (costs a full re-hash of the file), and append_loraname_if_empty adds the filename to both tag lists when they come back empty - a last-resort trigger guess.

Stacked, but not a loader

Remember what "stacked" means here: this node builds a stack, it doesn't apply it. There's no MODEL input or output. You chain stacks together and then hand the final LORA_STACK to something that consumes it - an efficiency-nodes stacked loader, or an unstack node that turns it back into individual LoRA pairs. New ComfyUI users routinely miss this and wonder why no model is being patched.

Installing

Part of ComfyUI-Lora-Auto-Trigger-Words. Install via ComfyUI Manager (search "ComfyUI-Lora-Auto-Trigger-Words") and restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/idrirap/ComfyUI-Lora-Auto-Trigger-Words

No extra dependencies, no model downloads, no network needed beyond the optional Civitai tag lookup. If the preview dropdown is a must-have, step up to the Advanced version - but this node is the honest recommendation for anyone who wants the functionality without the dependency.