Usage

• Try the LoRA on fal.ai Playground
• Try the workflow (image-to-video) at fal.ai Workflow
• ComfyUI: download compatible weights virtual-tryoff-lora_comfy.safetensors
• Diffusers:
  python

  1import torch 2from diffusers import Flux2KleinPipeline 3from PIL import Image 4 5pipeline = Flux2KleinPipeline.from_pretrained( 6 "black-forest-labs/FLUX.2-klein-base-9B", 7 torch_dtype=torch.bfloat16, 8 low_cpu_mem_usage=False 9).to("cuda") 10pipeline.load_lora_weights( 11 "fal/virtual-tryoff-lora", 12 weight_name="virtual-tryoff-lora_diffusers.safetensors", 13 adapter_name="vtoff" 14) 15pipeline.set_adapters("vtoff", adapter_weights=1.0) 16pipeline.fuse_lora(adapter_names=["vtoff"], lora_scale=1.0) 17 18image = pipeline( 19 image=Image.open("<your_image>.jpg"), 20 prompt="TRYOFF extract the full outfit over a white background, product photography style. NO HUMAN VISIBLE (the garments maintain their 3D form like an invisible mannequin).", 21 height=1024, 22 width=768, 23 num_inference_steps=28, 24 guidance_scale=5.0, 25 generator=torch.Generator("cuda").manual_seed(42), 26).images[0]

Training

• Base model: FLUX.2-klein-base-9B
• Steps: 10000
• Learning Rate: 0.00005
• Dataset: 300 image pairs (model + garment) of shape 1024x1024

Author