Downloads · 30 days
184
11% of all-time downloads
marioortegat/jewelry-product-photography-sdxl-lora
jewelry-product-photography-sdxl-lora is a text-to-image model from marioortegat. Use it when you need an image from a text prompt. It is set up for diffusers. The card lists the license as creativeml-openrail-m.
HiResJewels-SDXL is a high-resolution LoRA adapter for Stable Diffusion XL 1.0, fine-tuned specifically for luxury jewelry product photography. It excels in generating photorealistic macro shots, capturing complex lig…
Downloads · 30 days
184
11% of all-time downloads
All-time downloads
1.7K
Public
Repo size
457 MB
Likes
5
Public
Click a slice to open those files.
.safetensors457 MB · 100%
From the Hugging Face model README
HiResJewels-SDXL is a high-resolution LoRA adapter for Stable Diffusion XL 1.0, fine-tuned specifically for luxury jewelry product photography. It excels in generating photorealistic macro shots, capturing complex light reflections on metals, and rendering precise diamond settings.
This model was trained with a focus on high-fidelity textures and professional studio lighting conditions.
stabilityai/stable-diffusion-xl-base-1.0Use hiresjewels_style at the beginning of your prompt to activate the LoRA style.
sdxl_vae.safetensors (fp16 fixed) to ensure correct color saturation and avoid "washed out" images.You can run this model directly using the standard diffusers library:
import torch
from diffusers import StableDiffusionXLPipeline
# Load base model
pipe = StableDiffusionXLPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
torch_dtype=torch.float16
)
# Load LoRA weights
pipe.load_lora_weights("marioortegat/jewelry-product-photography-sdxl-lora", weight_name="pytorch_lora_weights.safetensors")
pipe.to("cuda")
# Inference
prompt = "hiresjewels_style, a diamond ring, macro shot, studio lighting, 8k"
image = pipe(prompt, num_inference_steps=30, guidance_scale=7.0).images[0]
image.save("jewelry_render.png")
Gold Texture & Studio Lighting
<div align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/690a5c5541ad780b8438fbc7/BfzJrL1Qi8-F8T2K_bNS_.png" width="600px" /> </div>Prompt:
hiresjewels_style, detailed product shot of a heavy gold chain necklace, 24k gold texture, studio softbox lighting, isolated on black background, professional photography, 8kNegative Prompt:(human skin:1.5), (hands:1.5), blurry, low quality, watermark, text
While this model excels at product visualization, users should be aware of the following limitations:
Text Rendering: The model may struggle with generating legible complex text or specific engravings on the metal.
Material Bias: The training dataset heavily featured platinum and rose gold; users may need to increase prompt weight to generate other materials (e.g., wood or resin) effectively.
Anatomy: This LoRA is optimized for product-only shots. Generating jewelry on human hands or necks may require additional negative prompting or ControlNet guidance to maintain anatomical correctness.