Load Lora Stacked with Previews (XTNodes)

The plain version of LoraLoaderStackedAdvancedWithPreviews - same idea of chaining local LoRAs into one LORA_STACK you apply downstream, minus the split between model and clip weight. One weight controls both here, which covers the large majority of real use: most of the time you're not trying to decouple how hard a LoRA pulls on the model from how hard it pulls on the text encoder, you just want it stronger or weaker overall.

How it works

Pick a local LoRA from the dropdown, set a weight, and this node appends (file, weight, weight) to whatever stack came in through the optional lora_stack input - chain several of these together and each one adds its own entry. Feed the result into an Apply LoRA Stack node to actually patch a model and clip with all of them at once; the README notes the stack shape is interoperable with Comfyroll's ComfyUI_Comfyroll_CustomNodes, so this pack's stackers and Comfyroll's applier can sit in the same workflow.

Same Civitai-lookup as the rest of the pack rides along: the file gets BLAKE3-hashed and checked against Civitai's database, and if it matches you get trigger words and preview images for free. If it doesn't - your own training, or something not on Civitai - the LoRA still stacks fine, you just get an empty trigger-word list instead of an error.

The inputs and outputs that matter

model_name is a dropdown of your local loras folder. lora_weight (default 1.0, -10 to 10) is the single weight applied to both model and clip. bypass (default off) flips this entry out of the stack without you needing to unplug it - the fast way to A/B a five-LoRA chain and see which one is actually doing the work. override_trigger_words and preview_images behave as in every loader here. lora_stack (optional) is where an upstream stacker chains in.

Output: LORA_STACK onward, plus civitai_trigger_words (LIST) and summary(Text).

How to install it

ComfyUI Manager, search "ComfyUI Easy Civitai (XTNodes)", or clone and pip install -r requirements.txt, then restart. This node is purely local - no aria2c, no API token needed here.

Common issues & troubleshooting

If you want the trigger-word-surfacing part of this without the Civitai integration, it's worth knowing rgthree-comfy's Power Lora Loader does a version of the same stacking job, is far more widely used, and is actively maintained - a reasonable alternative to have in mind.

If you're sticking with this pack: it hasn't had a commit since September 2024, and a subtle failure mode is baked into how it loads - __init__.py imports every node file and silently discards any exception one of them raises, no traceback, nothing printed. Two open GitHub issues report installing this pack and getting only its prompt-utility nodes, with every loader (this one included) simply missing - the likely cause is one of the five pip dependencies (blake3 especially, since it needs to compile on some systems) failing during install. If this node doesn't turn up after you install and restart, re-run pip install -r requirements.txt inside the exact environment ComfyUI runs in and read what it actually printed.