🎯 Bring ultimate sophistication to your generations with HighFashionXL, a premium LoRA designed to capture the essence of haute couture, luxury, and effortless elegance.
✨ What This LoRA Does Best:
The Signature Look: Perfectly renders an all-AI model exuding poise and class, styled in stunning, high-end designer dresses.
The Classy Touch: Highly optimized to naturally include elegant props like a wine glass, making it ideal for gala, lounge, or high-society aesthetics.
Premium Quality: Enhances textures, fabric drapes, lighting, and sophisticated poses without distorting your base model's face.
🎯 Trigger & Prompting Tips:
To get the best results, try incorporating these keywords into your prompt: high fashion, classy woman, elegant dress, holding a wine glass, luxury aesthetic, detailed fabric
Drop your generations in the gallery below—I'd love to see what you create!
FLUX IMPORTANT: Use 0.4-0.6 Strength or results will be over baked heavily!!
Flux 1 & 2 :
Sampler: DPM 2 Ancestral
CFG: 6.0
Steps: 20
Clip Skip: 2
SDXL:
Sampler: Euler
CFG: 7
Steps: 28-32
Clip Skip: 2
Qwen:
CFG: 6.5
Steps: 30-35
Clip Skip: 2
Upscaler: ESRGAN
Deploy this Model and much more:
Runpod
https://runpod.io?ref=1jcw9158
Generate 4K images and train Lora's
Run Diffusion
https://app.rundiffusion.com/?referralCode=63656285-d822-4b1e-9858-a7bb6c596d77
Krea 2 Raw & Other Premium Versions of this model:
https://www.patreon.com/TensorVizion/posts/highfashion-krea-167277892
Generative AI Workflows & Resources:
https://www.kofi.com/tensorvizion
Description
Trained on Z-Image
FAQ
Comments (1)
I tested the HighFashion-Qwen-Image_c1-st4000, weight 1.0, with qwen_image_21_int8_convrot and, unfortunately, couldn’t spot any difference in the clothing or the fabrics. Normally, I can spot significant differences between images with and without LoRa. In this case, however, there were only very slight changes to the fingernails or the width of the clothing. If this is down to the version, I look forward to an update with more noticeable results.
... but it could also be down to my LoRa node; qwen2.1 is still quite new...