🔧 LoRA for Flux Parameters

This one is for people who test Flux LoRAs systematically instead of eyeballing them. LorasForFluxParams packs one or more LoRAs, each with its own strength (or a list of strengths), into a single LORA_PARAMS bundle that cubiq's Flux parameter-sweep nodes then run through - generating a grid so you can compare LoRAs and strengths on the same seed and prompt side by side. If you've ever loaded a Flux LoRA at 1.0, decided it was too strong, changed it to 0.8, re-ran, and repeated that five times, this is the node that turns all of that into one execution.

It's from ComfyUI Essentials by cubiq (Matteo Spinelli, also the author of ComfyUI_IPAdapter_plus), and it's specifically part of the pack's Flux-testing tooling - not a general LoRA loader.

How it works

Instead of applying a LoRA to your model directly, it collects your LoRA choices and their strength values into a LORA_PARAMS object. The strength is a text field on purpose: you can type a single value like 1.0, or a list of values, so the downstream sweep node can iterate every strength you gave it. That's the whole trick - it defers the "apply" step so a sweep node can loop over the combinations.

The inputs and outputs that matter

• lora_1 (dropdown) - the LoRA to test. It reads from your models/loras folder, so if the dropdown is empty, you have no Flux LoRAs installed (or none in the expected folder).
• strength_model_1 (string, default 1.0) - the strength, as text so you can pass one value or a list to sweep.
• LORA_PARAMS (output) - the bundle. Wire it into cubiq's Flux sampler-parameters / sweep node; it isn't a MODEL you plug straight into a KSampler.

How to install it

ComfyUI Manager: search ComfyUI Essentials, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/cubiq/ComfyUI_essentials
pip install -r ComfyUI_essentials/requirements.txt

then restart. No models ship with the pack - you supply your own Flux LoRAs.

Common issues & troubleshooting

The LoRA dropdown is empty. The node lists files from your LoRA folder. If there's nothing there, drop your Flux LoRA .safetensors into ComfyUI/models/loras and refresh. Note this node's whole purpose is Flux - pointing it at SDXL LoRAs doesn't make sense.

I wired the output into a KSampler and it broke. LORA_PARAMS isn't a MODEL. It's an instruction bundle meant for the matching Flux parameter/sweep node in the same pack. Feed it there, not directly into sampling.

I wanted to just load one LoRA normally. Then you don't want this - use core ComfyUI's Load LoRA node. This exists specifically for structured comparison across LoRAs and strengths, which is overkill for a single fixed LoRA.

Missing after a Comfy update. Essentials is maintenance-only since April 2025, and pack nodes have broken against newer ComfyUI before. Update via Manager; a rollback is the usual fallback if it's genuinely incompatible.