XY Input: LoRA Plot

Most XY Input nodes drive one axis. This one drives both - it's the node behind the pack's signature demo, a full grid of LoRA model strength across the top and clip strength down the side (or a batch of LoRAs against a strength sweep). If you want a proper 2D matrix in one node instead of wiring up two separate XY inputs, this is it.

It feeds an XY Plot node; on its own it does nothing.

Why you'd reach for it

Finding a LoRA's best settings is a two-variable problem more often than people admit. A LoRA can have a strong model effect but overcook the CLIP side, so the interesting question isn't "what weight" but "what combination of model and clip strength." This node lays that out as a grid so you can spot the cell that actually looks right. It's also how you compare a whole folder of LoRAs while varying strength at the same time - batch on one axis, strength on the other.

The inputs and outputs

The key input is input_mode, which sets what the two axes are:

• X: LoRA Batch, Y: LoRA Weight
• X: LoRA Batch, Y: Model Strength
• X: LoRA Batch, Y: Clip Strength
• X: Model Strength, Y: Clip Strength

That last one is the classic model-vs-clip grid.

Then:

• lora_name - the LoRA to test (for the strength-vs-strength modes).
• model_strength / clip_strength - the baseline strengths.
• X_first_value / X_last_value and Y_first_value / Y_last_value - the ranges each axis sweeps.
• X_batch_count / Y_batch_count - how many steps (cells) each axis is divided into.
• X_batch_path / X_subdirectories / X_batch_sort - for the batch modes, the folder of LoRAs to sweep on the X axis.
• lora_stack (optional) - sit the plotted LoRA on top of an existing stack.

Unlike the other XY inputs, this node has two outputs, X and Y - wire both into the matching X and Y inputs of the XY Plot node. That's the part people miss.

Installing it

ComfyUI Manager → search Efficiency Nodes for ComfyUI, or:

cd ComfyUI/custom_nodes
git clone https://github.com/jags111/efficiency-nodes-comfyui

Restart after. LoRA files go in models/loras (or under your batch path).

Common issues

The most common wiring mistake is connecting only one of the two outputs. This node produces both X and Y; hook both into the XY Plot or you'll get a 1D strip instead of the 2D grid you set up.

Watch the total cell count too - X_batch_count times Y_batch_count is how many images it renders, and it climbs fast. A 6×6 is 36 generations; be deliberate before you queue a big one.

Same SDXL caveat as the other LoRA nodes: on the base+refiner path this pack has historically applied LoRAs to the base only, not the refiner, so weak-looking results there may be that quirk rather than your ranges. And it only works through an XY Plot feeding an Efficient KSampler. Pack-wide: a missing node usually means IMPORT FAILED at startup (often a pip freeze error) - update ComfyUI, the packages and the node.