LoraLoader_motorway_edition
Slap a LoRA onto both model and CLIP, pipe-style
- MOTORWAY ππ¨
- MOTORWAY ππ¨
LoRAs are the community's favorite way to steer a model - a small weight file that teaches a checkpoint a character, a style, or a concept. LoraLoader_motorway_edition is the full version of that loader: it pulls a model and a clip out of your Motorway pipe, applies the LoRA to both of them, and stores the patched pair back under new keys.
The reason the standard LoraLoader touches both is that a LoRA can patch the diffusion model and the text encoder (CLIP), and many LoRAs - character ones especially - encode part of their identity in the prompt side. You get two strength knobs for exactly this: strength_model controls how hard the LoRA pushes the diffusion model, and strength_clip controls the CLIP side. Both accept negative values, which is how people do "anti-LoRA" subtractive tricks. The tooltips in the schema spell it out: these control "how strongly to modify" each side, and they "can be negative."
The standard workflow here is: load your checkpoint somewhere upstream, write model and clip into the pipe, then chain a few of these nodes - one per LoRA - each reading the previous patch and writing the next. Because the pipe is cloned and updated as it flows, stacking three LoRAs is just three of these nodes in a row with different lora_name selections. That's the Motorway's whole pitch: no wire tangles, just keys.
The wrapper
MOTORWAY ππ¨- the pipe. Required.INPUT_model_key(defaultmodel) - where the base model lives.INPUT_clip_key(defaultclip) - where the CLIP lives.lora_name- dropdown of LoRA files from yourmodels/lorasfolder.strength_model(FLOAT, -100 to 100, default 1) - patch strength on the model.strength_clip(FLOAT, -100 to 100, default 1) - patch strength on the CLIP.OUTPUT_MODEL_key(defaultMODEL) andOUTPUT_CLIP_key(defaultCLIP) - where the patched pair lands.
Output port is the pipe; pull MODEL/CLIP out with a Motorway AxB ramp to hand them to a plain node (or just chain another LoraLoader_motorway_edition that reads MODEL/CLIP).
Installing
Part of the Motorway pack:
cd ComfyUI/custom_nodes
git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway
# restart ComfyUI
Or ComfyUI Manager β search ComfyUI_agilly1989_motorway β Install. No pack-specific model downloads - the LoRAs themselves are yours and live in models/loras.
Gotchas
- Start the graph with a
MotorwayStartand load the checkpoint into the pipe before this node runs -Key: 'model' doesn't exist in motorwaymeans the checkpoint never got written. - Keys are exact strings. If you rename an output key, every downstream reader has to use the new name.
- Negative strengths work but behave weirdly fast - the LoRA-training lore is full of "I set it to -1 and things got strange" stories. Start at 1.
- Beta pack: author's README says "CURRENTLY IN ACTIVE DEVELOPMENT (BETA)! IF THINGS BREAK ITS BECAUSE I BROKE IT," and the clone set has been toggled off in some releases. Missing from the menu after install? Check the pack's current version.
If you only need the model patched and not the CLIP, LoraLoaderModelOnly_motorway_edition is the lighter sibling. This one is for the full-strength character and style work.
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| MOTORWAY ππ¨ | MOTORWAY ππ¨ | β | |
| INPUT_model_key | STRING | model | β |
| INPUT_clip_key | STRING | clip | β |
| lora_name | COMBO | The name of the LoRA. | |
| strength_model | FLOAT | 1.00-100β100 | How strongly to modify the diffusion model. This value can be negative. |
| strength_clip | FLOAT | 1.00-100β100 | How strongly to modify the CLIP model. This value can be negative. |
| OUTPUT_MODEL_key | STRING | MODEL | β |
| OUTPUT_CLIP_key | STRING | CLIP | β |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| MOTORWAY ππ¨ | MOTORWAY ππ¨ | β |