load lora (menu)

Here's the twist: the most useful node in a LoRA-merging pack is the one that doesn't merge anything. load lora (menu) is a plain, honest LoRA loader - model in, LoRA applied, patched model out - with one feature the stock ComfyUI loader doesn't give you: a folder filter dropdown. If you keep LoRAs in subfolders, this is the node you'll reach for every single workflow, whether or not you ever touch the merge half of the pack.

What it actually does

Give it a MODEL (from a Load Checkpoint or any model loader), pick a LoRA, set a strength, and you get a patched model back. Under the hood it resolves the file from models/loras, loads the weights, and calls ComfyUI's own load_lora_for_models - the same machinery the built-in LoraLoader uses. The code even tries two calling conventions (state dict first, then file path) to survive different ComfyUI versions. It's not doing anything exotic, and that's the point: it's a dependable daily driver.

The strength_model slider runs from -2.0 to 2.0 with a 0.01 step, so you can go negative for a subtractive effect. Default 1.0. The old "0.5–0.8 is the sweet spot" advice mostly comes from the SDXL era; what's right depends on the LoRA, so start at the model page's recommendation and nudge from there.

The inputs and outputs

• model - the checkpoint model you're patching. Required.
• category_filter - folder dropdown (All + subfolders under loras). Filters the file list client-side via the pack's web script.
• lora_name - the actual file picker.
• strength_model - how hard the LoRA bites, −2 to 2, default 1.

Output: model, the patched model ready for a sampler. Same name in and out, which trips up newcomers - you're chaining it: load → patch → generate.

Installing

No models to download, no extra deps beyond safetensors (bundled with ComfyUI). Through ComfyUI Manager, search "comfyui-merge", or:

cd ComfyUI/custom_nodes
git clone https://github.com/LingSss9/comfyui-merge

Restart ComfyUI. One caveat from the source: if your environment is missing safetensors, .safetensors LoRAs fail with an import error - pip install safetensors into ComfyUI's env fixes it.

Honest downsides

The pack is a small side project - roughly a dozen GitHub stars, last touched in mid-2025 - so don't expect active development. The folder filter is cosmetic; it narrows the dropdown, not the search path. And if all you need is "load a LoRA with a strength slider," ComfyUI's own LoraLoader plus a folder-aware alternative like rgthree's Power LoRA Loader (which also reads trigger words off the file) covers the same ground with a bigger community behind it. Where this node wins is simple: it's right there in a pack you already installed for merging, and it does the job without ceremony.