ComfyUI-LoRAWeightAxisXY

XY Input: LoRA Weight (simple) is a tiny XY input node for the Efficiency Nodes that lets you sweep LoRA strength along one axis of an XY plot.
It outputs an XY value of subtype LoRA — a list of triples (lora_name, weight, weight) — linearly spaced from min → max in steps.
This makes it directly combinable with XY Input: Checkpoint (e.g., X = checkpoint, Y = LoRA weight) without having to use the more constrained “LoRA Plot” modes.

---

Why this node?

In Efficiency Nodes, the stock XY-plot logic blocks mixing Checkpoint with certain LoRA-special types ("LoRA Wt", "LoRA MStr", …) when only one axis is LoRA.
This node sidesteps that by emitting the regular LoRA subtype with the same strength applied to Model and Clip.
Result: you can render multiple checkpoints × multiple LoRA strengths in a single grid.

Typical use cases

• Compare how one LoRA behaves across different checkpoints.
• Find the best LoRA weight (0.0 → 1.0) for a specific checkpoint.
• Quickly explore LoRA intensity without the extra constraints of “LoRA Plot”.

---

Requirements

• ComfyUI
• Efficiency Nodes (XY Plot + Efficient Loader + KSampler)
• No additional dependencies

---

Installation

With ComfyUI Manager (recommended)

Use Install via Git URL, paste your repository URL, then restart ComfyUI.

Manual

Place the folder inside your ComfyUI directory and restart:

ComfyUI/custom_nodes/LoRAWeightAxisXY/
  ├─ __init__.py
  └─ lora_weight_axis_xy.py

---

Node overview

Name: XY Input: LoRA Weight (simple)
Category: Efficiency Nodes/XY Inputs

Inputs

• lora_name — dropdown from models/loras (pick the LoRA to sweep)
• min_value — start weight (default 0.0)
• max_value — end weight (default 1.0)
• steps — number of points including endpoints (default 3)
  → Example: 0.0 → 1.0, steps=3 → weights 0.00, 0.50, 1.00 (rounded to 2 decimals)

Output

• XY (label: X or Y), subtype LoRA, structure:

[
  [ (lora_name, weight, weight) ],
  [ (lora_name, weight, weight) ],
  ...
]

---

Quickstart (wiring)

1. Add Efficient Loader (checkpoint/CLIP/tokenizer as usual)
   → DEPENDENCIES → XY Plot.dependencies
2. X axis: XY Input: Checkpoint (e.g., two ckpt names) → XY Plot.X
3. Y axis: XY Input: LoRA Weight (simple) (choose your LoRA, e.g., min=0.0, max=1.0, steps=3) → XY Plot.Y
4. XY Plot.script → Efficiency Nodes / KSampler.script
   (must be the Efficiency KSampler with a script input)
5. In XY Plot, set ksampler_output_image = Plot (returns the grid as a single image)
6. KSampler.IMAGE → Save Image

Result: a grid sized (#checkpoints × #weight-steps) (e.g., 2×3).

---

Notes & limitations

• The node sets Model and Clip strength to the same value ((v, v)), which is sufficient for most “how strong?” comparisons.
  If you need separate model/clip axes, use the standard LoRA XY inputs — both axes must then be LoRA-special types.
• Avoid having a separate Load LoRA node with fixed weights active in your main path; it can override the sweep.
• Rendering must use the Efficiency KSampler (with the script input). Core/other KSamplers ignore XY plot scripts.

---

Troubleshooting

Only one image instead of a grid

• Confirm the KSampler is the Efficiency Nodes version (has a script input).
• Ensure XY Plot.dependencies is connected to Efficient Loader → DEPENDENCIES.
• Set ksampler_output_image = Plot in XY Plot.
• Check the terminal for an “XY Plot Error:” message (then recheck types/wiring).

Can’t connect the Y axis

• Some XY-plot variants show linkable ports only after clicking the small link/chain icon next to the field.

---

Examples

A) 2 checkpoints × 3 LoRA weights

• X: XY Input: Checkpoint → [ckpt_A, ckpt_B]
• Y: XY Input: LoRA Weight (simple) → LoRA=myStyle.safetensors, range 0.0 → 1.0, steps=3
  → 6 tiles: A×(0.00, 0.50, 1.00) and B×(0.00, 0.50, 1.00)

B) LoRA sweep only (single checkpoint)

• X: (empty)
• Y: XY Input: LoRA Weight (simple) → 0.0 → 0.8, steps=5
  → 1×5 strip

---

Files

• __init__.py
• lora_weight_axis_xy.py

---

License

MIT © 2025