LoRA Name List

LoRA workflows have a recurring annoyance: there's no standard way in ComfyUI to just list the LoRAs you want to use and hand that list to something else. You usually end up with one loader per LoRA, wired in a chain that's painful to edit. LoRA Name List attacks the selection half of that problem - it gives you dropdowns for up to twenty of your installed LoRAs and outputs them as a single Python list.

It doesn't load anything. It's the shopping list node. But a clean shopping list is the part most graphs lack.

How it works

At instantiation it scans your LoRA folder and builds the dropdown options:

loras = ["None"] + folder_paths.get_filename_list("loras")

That's why the choices show up automatically - any .safetensors in your models/loras directory appears in the dropdowns. When you run it, it walks the slots up to lora_count, skips anything still set to "None", and collects the rest into a list:

for i in range(1, lora_count + 1):
    lora_name = kwargs.get(f"lora_name_{i}")
    if lora_name and lora_name != "None":
        lora_list.append(lora_name)

A frontend extension (web/js/lora/name-list.js) hides the unused slots as you adjust lora_count, so the node stays compact instead of showing twenty dropdowns when you only use three.

Inputs and outputs

The inputs that matter:

• lora_count - how many slots to use, 1–20. Set this to 3 and only three dropdowns remain visible.
• lora_name_1 … lora_name_20 - the dropdowns. Each lists your installed LoRAs plus "None".

The one output:

• LIST - a Python list of the selected LoRA filenames, in the order you picked them, with "None" slots skipped.

Where the LIST goes

This is the part to think about before wiring it in: the output is a plain LIST type, and the consumers for that are generic - nodes that accept a list of strings. In the same pack, the LoRA Loader Stack is the natural companion: it takes LoRAs as individual dropdowns and applies them to a model in order. There's no built-in "feed the LIST into the stack" connection, so plan on this node feeding anything that expects a LIST/LIST[str] - batch processors, custom loaders, or a Python-eval style node if you're that kind of builder.

If your actual goal is "load several LoRAs onto a model in one node," you may not need this at all - go straight to LoRA Loader Stack, which does the selection and the loading. This node earns its keep when you need the names as data: logging, iteration, or driving a custom loader.

Install

Part of geocine-comfyui, one install gets all eleven nodes:

• ComfyUI Manager → search geocine-comfyui → install → restart
• or Comfy CLI: comfy node install geocine-comfyui
• or manually:

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

then restart ComfyUI. No model downloads, no extra pip packages (the pack only pulls openai, for the LLM node).

Gotchas

Your new LoRA isn't in the dropdowns. The options are read at node creation, so a freshly added LoRA may not appear until you re-add the node or restart ComfyUI. The LIST is filenames, not model objects - don't try to plug it into a MODEL input. And architecture lock-in applies as always: since the LoRA era began, a LoRA only runs on the model family it was trained for, so the names you pick still have to match your checkpoint.