ComfyUI Node

LoraLoaderAdvanced

The trigger-word LoRA loader with the fancy dropdown

By idrirap·Created 3 years ago·Updated about a year ago· 268
LoraLoaderAdvanced
  • model
  • clip
  • MODEL
  • CLIP
  • civitai_tags_list
  • meta_tags_list
  • lora_name
lora_name
strength_model1.00
strength_clip1.00
force_fetchfalse
append_loraname_if_emptyfalse
override_lora_name

LoraLoaderAdvanced is the flagship loader of the ComfyUI-Lora-Auto-Trigger-Words pack: everything LoraLoaderVanilla does - load a LoRA onto MODEL and CLIP, and emit its trigger words as two tag lists - plus a LoRA preview thumbnail in the dropdown and a right-click "View info..." dialog. It's the version you reach for when you're building a workflow around LoRAs and want to actually see what you're loading instead of trusting filenames.

The trade-off is spelled out in the name and in the docs: "Advanced" is a front-end feature, not a backend one. The loading logic is identical to the Vanilla node - under the hood LoraLoaderAdvanced is the same class - and the preview/info extras come from pythongosssss/ComfyUI-Custom-Scripts. The pack literally reuses that project's betterCombos.js for the preview. So the decision between Advanced and Vanilla is really: do you want thumbnails and the info panel, and are you willing to install Custom-Scripts to get them?

How it works

Same inputs as the Vanilla loader: model (MODEL), lora_name (dropdown), strength_model and strength_clip (both default 1.0), plus force_fetch and append_loraname_if_empty booleans, and optional clip and override_lora_name. Same five outputs: patched MODEL and CLIP, civitai_tags_list and meta_tags_list (the trigger-word lists), and lora_name.

The tag mechanism is unchanged and worth repeating because it's the reason this pack exists: civitai_tags_list hashes the file and asks Civitai's public API for trainedWords, cached in loras_tags.json; meta_tags_list reads ss_tag_frequency straight out of the safetensors header and sorts by training frequency. No API key, no config. The first fetch on a big LoRA takes a moment because it hashes the whole file; afterwards it's cached.

The preview and "View info..." are where the Custom-Scripts dependency bites. If Custom-Scripts isn't installed, the dropdown is plain - and at that point the Vanilla node is functionally identical, so don't bother installing this pack twice. For the right-click info dialog, enable it under Settings → Pysssss → Model Info → Model Info - Lora Nodes/Widgets and append LoraLoaderAdvanced.lora_name to the widget list.

Where people get burned

The common mistake is expecting the trigger words to magically appear in your prompt. They won't - this loader emits the lists, and you still have to wire them through TagsFormater (to see them with indexes) and TagsSelector (to filter and weight them) before they reach your CLIP Text Encode. The other gotcha is the dependency: if your workflow loads with a missing-nodes error after a fresh install, check that Custom-Scripts is present - it's the most common reason this node renders without its features.

Installing

Install the pack via ComfyUI Manager (search "ComfyUI-Lora-Auto-Trigger-Words") and restart, then install ComfyUI-Custom-Scripts (pythongosssss) the same way:

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

No model downloads, no heavy Python deps - the Civitai lookup runs on requests, which ComfyUI already ships. If you don't want the Custom-Scripts dependency at all, LoraLoaderVanilla gives you the same loading and the same tag lists without the preview.

Categoryautotrigger

Inputs (8)

NameTypeDefaultDescription
modelMODEL
lora_nameCOMBO0 options:
strength_modelFLOAT1.00-10–10
strength_clipFLOAT1.00-10–10
force_fetchBOOLEANfalse
append_loraname_if_emptyBOOLEANfalse
clipoptCLIP
override_lora_nameoptSTRING

Outputs (5)

NameTypeDescription
MODELMODEL
CLIPCLIP
civitai_tags_listLIST
meta_tags_listLIST
lora_nameSTRING