dimmexer/flux-device-mockup-lora

FLUX.1-dev Device Mockup LoRA

A specialized LoRA fine-tune for FLUX.1-dev designed to generate high-quality, photorealistic device mockups—including smartphones, laptops, tablets, and desktop displays—featuring clean, blank white screens ideal for product showcases, UI/UX presentations, and design mockups.

Model Details

• Trigger Word: devicemk
• Base Model: black-forest-labs/FLUX.1-dev
• Model Type: LoRA (Low-Rank Adaptation)
• Rank / Alpha: 16 / 16
• Training Epochs / Steps: 1500 steps
• Training Framework: ai-toolkit
• Hardware: NVIDIA A100-SXM4-40GB

Recommended Generation Settings

• LoRA Strength / Weight: 0.8 - 1.0 (optimal at 1.0)
• Guidance Scale (CFG): 3.5
• Inference Steps: 28 - 35
• Sampler / Scheduler: FlowMatch / Euler
• Resolution: 1024x768, 1024x1024, or 768x1024

Trigger Word & Prompting Tips

Always include the trigger word devicemk in your prompt. Describe the device type, environment, lighting, and specify "blank pure white screen" for seamless UI insertion.

Example Prompts

1. Smartphone on Velvet Cushion:

   devicemk, a sleek modern smartphone on an orange velvet sofa cushion, blank pure white screen, soft studio lighting, photorealistic 8k

2. Laptop in Brutalist Setting:

   devicemk, an open modern laptop resting in a brutalist concrete stone pillar cutout, blank pure white screen, dramatic studio lighting

3. Handheld Tablet:

   devicemk, a modern tablet ipad held in hands, blank pure white screen, clean studio background

4. Floating Pastel Aesthetic:

   devicemk, a minimalist smartphone floating above a pastel podium, blank screen, subtle soft shadows, clean aesthetic, 8k

5. Architectural Studio Desk:

   devicemk, an ultra-thin laptop and dual monitor desk setup in an architectural design studio, blank pure white screens, soft natural window light, photorealistic

How to Use with Diffusers

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16)
pipe.load_lora_weights("dimmexer/flux-device-mockup-lora")
pipe.to("cuda")

prompt = "devicemk, a sleek modern smartphone on an orange velvet sofa cushion, blank pure white screen, soft studio lighting, photorealistic 8k"
image = pipe(prompt=prompt, num_inference_steps=28, guidance_scale=3.5).images[0]
image.save("device_mockup_output.png")

License

This model is built on top of FLUX.1-dev and follows the FLUX.1-dev non-commercial license.