Nodes/AUN ComfyUI Nodes/LoRA Loader Model Only (String)
ComfyUI Node

LoRA Loader Model Only (String)

Feed a LoRA name from another node — no dropdown required

By loz2754·Created 8 months ago·Updated a day ago· 5
LoRA Loader Model Only (String)
  • model
  • MODEL
  • resolved_lora
lora_name
strength_model1.00

Here's a niche that only makes sense once you've hit it: you want to apply a LoRA whose name comes from somewhere else in the graph - a random LoRA picker, a text node, an index switch - rather than from a dropdown you set by hand. ComfyUI's core LoraLoaderModelOnly only accepts a lora_name from its static combo list, which means you literally can't wire a dynamic name into it. AUNLoraLoaderModelOnlyFromString exists to fix exactly that: it takes the LoRA name as a plain STRING input, loads it onto your MODEL, and works anywhere a string can come from.

How it works

Mechanically it's a thin wrapper around the model-only LoRA load: give it a model, a lora_name string, and a strength_model, and it returns the patched MODEL. The name you feed it is a LoRA filename or relative path under your loras directory - the tooltip's example is SDXL/my_lora.safetensors, so subfolder paths work. That's the whole node: three inputs in, two outputs out.

The second output, resolved_lora, is the quietly useful one: it returns the resolved LoRA name/path after loading. Wire it into a filename builder or a text display and your saved outputs can record exactly which LoRA was applied - which matters a lot when the name was chosen dynamically and you'd otherwise have no record.

This is the kind of node that sits quietly in the LoRA-crafter's toolkit next to the stack-style loaders. If you mostly apply LoRAs by hand from a dropdown, you don't need it. If you're building anything that selects LoRAs programmatically - random loaders, prompt-indexed stacks - this is the adapter that makes it work with the standard model-only path.

Inputs and outputs

Inputs: model (MODEL), lora_name (STRING, the filename or relative path), strength_model (FLOAT).

Outputs: MODEL (patched), resolved_lora (STRING).

Installing it

Part of the AUN pack:

  • ComfyUI Manager: search "AUN ComfyUI Nodes", install, restart.
  • Manual: cd custom_nodes && git clone https://github.com/loz2754/AUN-ComfyUI-Nodes, then restart.

Hand installs hitting ModuleNotFoundError: cv2? pip install -r custom_nodes/AUN-ComfyUI-Nodes/requirements.txt.

Common issues

The main failure is feeding a name that doesn't exist - since there's no dropdown validating your input, a typo or a path that doesn't exist under loras means a load error or a silently empty patch. Double-check the string you're feeding. Also note this is model-only: it patches MODEL but not CLIP, so if you're applying a LoRA that also needs a CLIP-side strength (many do for text encoding), you want the model+clip variants in this pack, not this one.

CategoryAUN Nodes/Loras

Inputs (3)

NameTypeDefaultDescription
modelMODEL
lora_nameSTRINGLoRA filename or relative path under the loras directory (for example SDXL/my_lora.safetensors).
strength_modelFLOAT1.00-20–20LoRA strength applied to the model.

Outputs (2)

NameTypeDescription
MODELMODEL
resolved_loraSTRING