TriggerWord Toggle (LoraManager)
Switch individual trigger words on and off
- filtered_trigger_words
Loading a LoRA through this pack gives you its trigger words automatically - which is great until you've got five LoRAs stacked and their words are fighting each other for control of your prompt. TriggerWord Toggle is the node that lets you keep some and drop others without editing text by hand. It sits between a loader's trigger_words output and your prompt, and turns that pile of words into clickable on/off toggles right on the canvas.
That's a real workflow problem. A LoRA's trigger word is often mandatory - it's the first thing on the "why is my LoRA doing nothing" checklist - but the trigger words that came bundled with a LoRA sometimes include style tags you don't want this run. Rather than deleting them from a prompt string and remembering to put them back, you flip a toggle.
How it works
You feed in the trigger words coming out of a loader or stacker, and the node renders them as a set of switches. Whatever's toggled on gets passed through as filtered_trigger_words; whatever's off is dropped. Three settings shape the UI and behavior.
Inputs and outputs that matter
The three you actually set are all booleans:
group_mode(default true) - group the toggles by which LoRA they came from, so you can reason about them per-LoRA instead of one flat list.default_active(default true) - new trigger words start switched on. Leave it true if you usually want everything, flip it false if you'd rather opt words in.allow_strength_adjustment(default false) - turn this on to weight individual words (the(word:1.2)emphasis form) instead of a plain on/off.
One output: filtered_trigger_words (STRING) - the surviving words, wired into your CLIP encode or a Prompt (LoraManager) node.
How to install it
Via ComfyUI Manager: search lora-manager, Install, restart. Or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/willmiao/ComfyUI-Lora-Manager.git
cd ComfyUI-Lora-Manager
pip install -r requirements.txt
then restart. Light install - no models, no GPU dependencies.
Common issues & troubleshooting
No toggles appear. The node needs trigger words coming in from a loader or stacker. If that upstream trigger_words output is empty - because the manager never fetched words for those LoRAs - there's nothing to toggle. Set the words in the manager UI first.
Custom-widget rendering glitches. This node draws its own interactive UI, and interactive custom widgets are exactly the class of node that ComfyUI's Nodes 2.0 frontend rewrite has given trouble across the ecosystem (rgthree's toggles hit the same wall). If the toggles render stacked or won't respond, try resizing the node, and if you're on the new frontend, the standing community workaround has been to fall back to the legacy canvas.
Words don't reach the image. Toggling on isn't enough - the filtered_trigger_words output still has to be wired into the prompt that actually gets encoded. And the LoRA itself has to match your base model, or its trigger word does nothing regardless.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| group_mode | BOOLEAN | true | When enabled, treats each group of trigger words as a single toggleable unit. |
| default_active | BOOLEAN | true | Sets the default initial state (active or inactive) when trigger words are added. |
| allow_strength_adjustment | BOOLEAN | false | Enable mouse wheel adjustment of each trigger word's strength. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| filtered_trigger_words | STRING | — |