Put LoRA files in ComfyUI/models/loras. Then add a Load LoRA node to your workflow and pick the file in its lora_name list.
Wiring it up
The Load LoRA node takes a model and a clip input and returns a modified model and clip. Connect the checkpoint loader's MODEL and CLIP outputs into it, and the node's outputs on to the rest of the graph in their place. Its description says several LoRA nodes can be linked together.
Two inputs control the strength: strength_model (how strongly to modify the diffusion model) and strength_clip (how strongly to modify the CLIP model). Both default to 1.0 and can be negative. If both are 0 the node passes the model through unchanged.
The folder accepts the same file extensions as checkpoints (.safetensors, .ckpt, .pt and others); press R to refresh after adding one.
See where checkpoints go for the other model folders. If a LoRA comes from a custom node pack, the ComfyUI custom node install guide explains how those installs work.
Sources
- ComfyUI nodes.py (LoraLoader). github.com
- ComfyUI folder_paths.py. github.com