List: LoRAs (lab)

List: LoRAs (lab) is the LoRA-flavored twin of List: Checkpoints (lab): it gives you a multi-select of everything in your models/loras folder and outputs your picks as a LIST, ready to drive a queue. You use it when the question is "which of these six LoRAs actually makes this prompt work?" - or which epoch, which weight, which merge - and you want one batch run instead of six manual rebuilds.

The selection widget is populated from folder_paths.get_filename_list('loras'), the same source ComfyUI's standard LoraLoader dropdown reads. Snapshot at load time, so a freshly-downloaded LoRA needs a workflow reload before it appears.

Inputs and outputs

• selection - check the LoRAs you want in the comparison.
• with_extension - keep .safetensors on the names (default) or strip to the bare stem.

Outputs are the standard pair:

• list - the selected LoRA filenames, in your selection order.
• count - INT, how many you picked.

The real work happens downstream. This list wants to go into a queue node that feeds a loader's LoRA-name input. Because ComfyLab's queue outputs are untyped (Any), the name flows into a standard LoraLoader's filename input and the loader does the heavy lifting - stacking it on your checkpoint.

The pattern that makes it worth it

Hook it to XY Plot: Queue (lab) on dim1, with epochs or weights or seeds on dim2. The queue iterates every dim2 value for each dim1 value before advancing, and loading a LoRA is the slow operation - so it belongs on dim1, loaded once and cached while everything else varies. The KB's LoRA training material is full of exactly these grids: epochs across one axis, seeds down the other, one glance answers "which training run generalizes."

If you just want to stack a fixed set of LoRAs every run, this isn't the node - that's LoraLoader (or a power loader from another pack). ListLoras is for varying, not stacking.

Install

ComfyLab Pack is one install covering its whole node family (look for the "(lab)" suffix in node menus). ComfyUI Manager: search ComfyLab Pack, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/bugltd/ComfyLab-Pack.git
cd ComfyLab-Pack
pip install -r requirements.txt

Restart after. No model files - the pack is pure tooling, and its requirements (opencv-python, pyyaml, jsonschema and a few small ones) are light.

Gotchas

• Stale list - LoRAs downloaded after the workflow loaded won't show up until you reload.
• Extension consistency - keep with_extension matching what your loader wants; a mixed list is a silent failure waiting to happen.
• Name ≠ identity - if two LoRAs share a stem in different subfolders, with_extension off can hand back ambiguous names. Keep extensions on if you use subfolder organization.

The node is boring in the best way - a checkbox list that turns into a batch. For anyone who tests LoRAs against the same prompt repeatedly, it kills a surprising amount of busywork.