Dynamic LORA Selector

If you've used rgthree's Power Lora Loader - one of the most-borrowed nodes in the whole ComfyUI ecosystem, by corpus mentions - this node will look familiar in spirit: pick multiple LoRAs, give each one a weight, done. The important difference is what it's for. Power Lora Loader actually applies the LoRAs to your model and CLIP. This node doesn't load or apply anything - it builds a description of a LoRA combination, as data, for something else in your workflow to consume. That makes it a fit for the comparison-run pattern this whole pack is built around, not a replacement for your actual LoRA loading step.

What it does

Ten paired slots: lora_1 through lora_10, each an enum dropdown defaulting to none, and a matching weight_1 through weight_10, each a FLOAT defaulting to 1 with a range of 0 to 10. Pick the LoRAs you want in your combination, dial in a weight for each, and leave the rest on none. The node filters out the unused slots and returns your active picks two ways: a formatted STRING (name:weight pairs, comma-separated) and a LIST of the same data - the same aggregation pattern as this pack's Sampler and Scheduler generator nodes, just for LoRA stacks instead.

One detail worth flagging honestly: each dropdown offers none and a second option called Empty_value alongside your actual installed LoRAs. The schema doesn't document what distinguishes the two - both appear to represent "nothing selected here," possibly in slightly different ways for whatever's consuming the output downstream. If that distinction matters to your workflow, it's worth testing both rather than assuming.

Installing it

ComfyUI Manager: search DemonAlone-nodes-ComfyUI, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/DemonAlone/SimpeStringGenerator_ComfyUI

No dependencies or downloads bundled - you'll need your own LoRA files already in place for the dropdowns to list anything.

Where people get burned

The weight range goes all the way up to 10, which is far beyond sane LoRA territory. A LoRA's default weight of 1.0 is already frequently described as too strong by trainers - 0.5 to 0.8 is the common working range for most LoRAs. Cranking a slot into the 2-10 range isn't a stronger version of the same effect, it's very likely to break the image outright unless you specifically know what you're doing and why.

And remember the output-shape caveat that applies to every "generator" node in this pack: what comes out is names and weights as text/list data, not applied model patches. You still need an actual LoRA-loading node per entry to turn a selection into something that changes your generation - this node tells you what to load, it doesn't load it.