SLP Lora Loader (Model Only)

Some LoRA workflows don't have a CLIP half. Flux, for instance, commonly applies LoRAs to the diffusion model only - there's no CLIP in the pipeline to patch, and dragging a clip input around just to satisfy a loader is dead weight. SLP Lora Loader (Model Only) is the trimmed sibling of the pack's SLP Lora Loader: same text-input design, but the CLIP port is gone. Model in, model out, plus the handy stem string.

It exists because the pack author wanted the one-click LoRA switcher to fit both worlds - full SD/SDXL graphs with CLIP, and model-only graphs (Flux being the big one) without it. If you never connect a CLIP wire, this is the version you want; the full loader's CLIP port would just sit there empty.

How it works

It doesn't reinvent anything. The class literally subclasses SLPLoraLoader and calls the same underlying load_lora, passing clip=None and strength_clip=0. So every behavior carries over: only the first line of lora_name_text is used, an empty name is a clean passthrough, and the last-loaded LoRA is cached so repeated runs don't re-read the file.

Inputs and outputs

• model (MODEL) - the diffusion model being patched.
• lora_name_text (STRING) - the LoRA name as it appears in your loras folder, subfolders included. Feed it from SLP List View's selection for click-to-swap, or from SLP Directory Contents.
• strength_model (FLOAT, default 1.0, range −100 to 100) - how strongly the LoRA modifies the model. Negative values invert. As with any LoRA, 1.0 is often too strong; 0.5–0.8 is a common starting point.
• Outputs: MODEL (the patched model) and stem (STRING) - the LoRA filename without its extension.

stem is the output that makes this pack more than a novelty: run it into SLP Filename Prefix's $1 and every saved image is tagged with the LoRA that produced it. On Flux especially, where a single LoRA is often the whole style identity of a run, that filename stamp is genuinely useful.

Notes and gotchas

Because it passes clip=None into the shared loader, only LoRA files that target the model weights will apply; anything whose LoRA keys expect a CLIP side is silently ignored rather than erroring. That's expected for this node, not a bug. The other failure modes are the standard LoRA ones - wrong base model family, missing trigger word, or a filename typo in lora_name_text. And remember the picker pattern: before you click a row, SLP List View outputs an empty string, which the loader treats as passthrough - so a "LoRA not applying" mystery usually means no selection, not a bad file.

Install

One pack, one install, all five nodes. ComfyUI Manager (search ComfyUI-SingleLinePicker) or:

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

Restart ComfyUI afterward. No pip dependencies, no model files to download.

Use the full SLP Lora Loader when your graph has CLIP and you want the LoRA to patch both halves; reach for this one when it doesn't. Most days, if you're on Flux, this is the one you want.