🌊hua_gradio_Lora Model Only

Hua_LoraLoaderModelOnly (🌊hua_gradio_Lora Model Only) is the LoRA node that actually surfaces in the ComfyUI_to_webui frontend. Drop one of these into your workflow and the Gradio UI renders a LoRA dropdown for it; the user picks a LoRA and a strength, and on the next run the frontend writes their choice back into this node. It's the dynamic LoRA component the README advertises - one dropdown per instance, up to the configurable cap - and it's the single most useful node in the pack for anyone packaging a character- or style-consistent workflow.

Mechanically it's a stripped-down LoraLoader: it inherits the full loader's logic but takes MODEL in and returns a single MODEL out. No CLIP in or out - "ModelOnly." Why does that matter? Because the name you give each instance is what the webui uses to label its dropdown. Add three of these with sensible names and your packaged UI shows three clearly-named LoRA pickers, each feeding its own style into the graph. That's the "dynamic component" behavior working as intended.

Inputs:

• model - MODEL from your checkpoint/UNet loader.
• lora_name - dropdown from models/loras; the user picks here.
• strength_model - FLOAT, default 1.0, range −100 to 100. This is the only strength control.
• name - the label the webui shows above the dropdown. Defaults to "Hua_LoraLoaderModelOnly" - change it.

Output: MODEL. Chain it into your sampler (or into another LoRA node to stack).

Install: ComfyUI Manager → "ComfyUI_to_webui", or

cd ComfyUI/custom_nodes
git clone https://github.com/kungful/ComfyUI_to_webui.git

restart, and put your LoRAs in models/loras - nothing else to download. If a dropdown ever shows empty in the webui, the LoRA list is stale; there's a "refresh models" button for that.

The nuance that trips people: the "None" option. The frontend prepends None to the LoRA list, and when the user leaves it on None the node doesn't update lora_name - so the workflow keeps whatever LoRA the saved JSON had. If you want a truly neutral default (LoRA off until chosen), set the node's lora_name to a value and understand the user's "None" choice won't reset it. It's a small design quirk, not a bug, but it surprises people the first time. If you only need a fixed LoRA with no user choice, use the full Hua_LoraLoader instead - it stays out of the UI entirely.