LoRA Enable With Prefix

LoRA Enable With Prefix is a LoraLoader with a switch and a text combiner bolted on. In one node you get the standard LoRA loading - model and CLIP in, lora_name, strengths - plus a boolean enable toggle, plus a prefix and text pair that get merged into a single prompt string on the way out. The three outputs are the modified model, the modified CLIP, and the assembled text.

The reason this exists is that comparing "with LoRA" vs "without LoRA" in ComfyUI normally means ripping out wires or building two parallel sampler branches. This node lets you flip one checkbox and re-run the exact same graph. Toggle enable off and the LoRA simply isn't applied - you get your base model back, with the text output collapsing to just the prefix. That's the whole workflow: one boolean, two runs, clean A/B comparison of whether a style or character LoRA actually helps your current prompt.

How it works

Under the hood it's not reimplementing anything - it instantiates ComfyUI's own LoraLoader and calls load_lora with your lora_name and strengths. So it behaves exactly like the core node: same LoRA list (it reads from your models/loras folder), same strength ranges, same model/CLIP patching. If you've already got a LoRA that works in a plain LoraLoader, it'll work here identically.

The text logic is simple and worth internalizing because it affects what lands in your prompt:

• enable off → final text is just prefix.
• enable on → final text is prefix + text (space-joined), or whichever one is non-empty.

So the pattern is: prefix carries whatever should always be in the prompt (a character name, a base style), and text carries the thing tied to the LoRA (the LoRA's trigger words). Flip the toggle and the trigger words disappear along with the LoRA - which is exactly what you want for a fair comparison, because a trigger phrase with no LoRA behind it does nothing useful anyway.

Inputs and outputs

The important ones:

• enable (BOOLEAN) - the kill switch. Default on.
• lora_name - dropdown of your installed LoRAs.
• strength_model / strength_clip (FLOAT, −5 to 5) - the usual LoRA strength dials.
• text (multiline STRING) - the LoRA trigger words, typically.
• prefix (STRING, force-input socket) - always-on prompt text.

Outputs: model (MODEL), clip (CLIP), text (STRING). Wire model and clip forward like any LoRA node; wire text into your prompt build.

Installing it

Part of ComfyUI Fictiverse Nodes. ComfyUI Manager → search "ComfyUI Fictiverse Nodes" → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Fictiverse/ComfyUI_Fictiverse

No extra dependencies - it uses ComfyUI's built-in LoRA loader - and no model downloads. Apache 2.0.

Common issues

Pack startup gotcha first: ModuleNotFoundError: No module named 'custom_nodes.ComfyUI_Fictiverse' means the folder isn't named exactly ComfyUI_Fictiverse; rename and restart.

For this node specifically: the prefix input is force-input, meaning it only takes a wire - you can't just type into it. Leave it unwired if you don't need it. And watch the strengths: negative strengths are allowed (−5 to 5) and are occasionally useful, but they can also produce garbage quickly. If you're new to LoRAs, keep strengths at 1 and only tune strength_model - lowering strength_clip hurts prompt adherence more than people expect.