SLP Lora Loader

ComfyUI's built-in LoraLoader is fine until you want to swap LoRAs mid-batch, at which point its combo box becomes a productivity tax: click, scroll, click, run, repeat. SLP Lora Loader is the same loader with one deliberate difference - the LoRA name comes from a text input instead of a dropdown. Which means another node can feed it. Which means SLP List View can feed it, and you're back to the one-click LoRA switching the whole SingleLinePicker pack exists for.

LoRAs themselves are the standard SD adapter: small files that patch a checkpoint's cross-attention layers to add a character, style, or concept without replacing the base model. This node doesn't change any of that math - it changes how you select the file, and adds one genuinely useful output on the way out.

How it works

Mechanically it's near-identical to the stock loader: load the torch file, call comfy.sd.load_lora_for_models with the model and CLIP, apply the strengths. A few deliberate tweaks:

• lora_name_text is a plain STRING, and only its first line is used - so feeding it a multi-line selection (say, from a picker) still resolves to the top line. Nice safety net.
• Empty text, or both strengths at zero, returns the inputs untouched - a clean passthrough instead of an error.
• It caches the last loaded LoRA. Run the graph again with the same selection and it skips re-reading the file from disk, which matters when you're iterating.

Inputs and outputs that matter

• model (MODEL) and clip (CLIP) - the checkpoint pipeline the LoRA patches, exactly like the stock loader.
• lora_name_text (STRING) - the name as it appears in your loras folder, including subfolders (style/foobar.safetensors). It must match ComfyUI's own file list, which is why "Get Lora List" on SLP List View is the natural companion - it pulls those exact names.
• strength_model and strength_clip (FLOAT, default 1.0, range −100 to 100) - how hard the LoRA pushes on each half. Same rules as anywhere else: 1.0 is often too strong, 0.5–0.8 is a common starting point, and negatives invert.
• Outputs: MODEL and CLIP (the patched models, wired on as usual) plus stem - the LoRA filename without its extension. That third output is the gift: feed stem into the pack's SLP Filename Prefix and your saved images get the LoRA name baked into the filename.

When it bites

The usual LoRA failure modes still apply - wrong base model family (an Illustrious LoRA won't work on Flux), a missing trigger word, or a filename typo. The last one is the easy trap here: the loader resolves lora_name_text against folder_paths, so an exact filename mismatch fails. If your picker list came from "Get Lora List" or SLP Directory Contents you're safe; if you typed it, check the extension and subfolder. And remember the picker outputs an empty string before you've clicked a row - a blank lora_name_text passes through silently rather than erroring, so a "why is my LoRA not applying" moment usually means the picker never got a selection.

Install

Part of the hetima/ComfyUI-SingleLinePicker pack - one install gets all five nodes. ComfyUI Manager (search ComfyUI-SingleLinePicker) or:

cd ComfyUI/custom_nodes
git clone https://github.com/hetima/ComfyUI-SingleLinePicker

Restart ComfyUI. No pip packages, no model downloads.