Nodes/WAS Node Suite v3/Lora Input Switch
ComfyUI Node Runs on cloud

Lora Input Switch

Toggle between two LoRAs in ComfyUI

By WASasquatch·Created 3 years ago·Updated a day ago· 1,843
Lora Input Switch
  • model_a
  • clip_a
  • model_b
  • clip_b
  • MODEL
  • CLIP
booleantrue

Lora Input Switch picks one of two LoRA inputs based on a boolean, and passes it on. That's it - a two-way selector for the LoRA path in your graph. Flip the boolean and you swap which LoRA feeds downstream, without rewiring anything. It's part of WAS Node Suite's family of "Input Switch" nodes, which all follow the same pattern: two inputs, a switch, one output.

Why bother? Because it lets you A/B without dragging cables around. Set up two LoRA branches, wire both into the switch, and toggle between them from one control. Great for comparing two style LoRAs on the same seed, or building a workflow where a single flag decides which LoRA (or LoRA branch) is active on a given run.

How it works

The node takes two LoRA-carrying inputs - call them A and B - plus a boolean that decides which one wins. When the boolean is one value it forwards A; when it's the other it forwards B. Whatever's on the losing branch is simply ignored for that run. The chosen input flows out the single output and continues down your graph.

This is the same switch logic as WAS's Image Input Switch, Text Input Switch, and the rest - the family is built so you can gate almost any data type on one boolean flag.

The inputs and outputs that matter

The three connections are the two LoRA inputs you're choosing between and the boolean that selects. The output is the selected LoRA input, passed through unchanged. Drive the boolean from a Logic Boolean Primitive toggle (manual A/B) or from any boolean-producing node if you want the choice made automatically.

How to install it

Bundled in WAS Node Suite. Install once via ComfyUI Manager (search WAS Node Suite, install, restart) or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui
pip install -r was-node-suite-comfyui/requirements.txt

then restart. On Windows portable, run the pip step with python_embeded\python.exe -s -m pip install -r ... (or install.bat) so requirements install into ComfyUI's Python. No models of its own.

Common issues & troubleshooting

Both branches run even though only one is chosen. ComfyUI evaluates what it needs, and depending on how your branches are built, an unselected upstream branch can still execute. If a heavy LoRA-load branch you're not using is costing you time, that's the usual cause - restructure so the expensive work sits behind the switch where it can be skipped, not in front of it.

Switch direction feels backwards. Which boolean value picks A vs B is easy to get inverted. If the toggle selects the opposite of what you expect, just flip the boolean rather than rewiring the two inputs. Confirm with a quick A/B and label it in your head.

You wanted to stack LoRAs, not choose one. This node selects between two - it doesn't merge them. If your goal is running multiple LoRAs together with per-entry toggles, this is the wrong tool; a stacking loader (rgthree's Power LoRA Loader is the community favorite) is what you want. Use the switch for either/or, the stacker for both-at-once.

The suite won't import. WAS Node Suite is large and unmaintained since late 2023, and its trademark failure is an "Import Failed" after a ComfyUI update, from a dependency version clash (opencv especially). Reinstall the requirements against your ComfyUI Python and restart. The switch node itself is simple and stable - the suite install is the fragile part.

CategoryWAS Suite/Logic/Switch

Inputs (5)

NameTypeDefaultDescription
model_aMODELThe diffusion model sent on when boolean is true, with clip_a.
clip_aCLIPThe text encoder sent on when boolean is true. Wire it from the same LoRA loader as model_a, so a patched model keeps the CLIP it was patched with.
model_bMODELThe diffusion model sent on when boolean is false, with clip_b.
clip_bCLIPThe text encoder sent on when boolean is false. Wire it from the same LoRA loader as model_b.
booleanBOOLEANtrueWhich pair passes; BOOLEAN. true = model_a and clip_a, false = model_b and clip_b. Toggle it, or wire it from Logic Boolean.

Outputs (2)

NameTypeDescription
MODELMODELThe model half of the selected pair.
CLIPCLIPThe text encoder half of the same selected pair.