You can deploy this model and start earning money today!

shivam909067/
sdxl-bottle-marketing-lora

Name: sdxl-bottle-marketing-lora
Availability: InStock

apache-2.0

Views

No views yet

library_name: diffusers base_model: stabilityai/stable-diffusion-xl-base-1.0 tags:

lora
sdxl
product-photography
marketing
bottle
packaging
commercial
text-to-image
image-to-image license: openrail++ language:
en

SDXL Bottle Marketing LoRA

Professional Marketing-Quality Bottle & Packaging Photography Generator

📸 Model Description

A specialized SDXL LoRA fine-tuned on 641 professional bottle and packaging images for generating high-quality marketing photography. Trained for 2800 steps with enhanced LoRA architecture (dim=128, alpha=64) to achieve sharp edges, realistic geometry, and commercial-grade quality.

🖼️ Example Outputs

Luxury Glass	Essential Oil	Modern Cosmetic

Organic Green	Premium Clear	Purple Serum	Black Luxury

All images generated at 1024×1024 with CFG 5.5, 32 steps, LoRA weight 0.7

🎯 Optimized For

Glass bottles (clear, colored, frosted, amber)
Cosmetic & skincare packaging
Essential oils & wellness products
Beverage bottles
Luxury & premium packaging
Minimalist modern designs
Natural & organic aesthetics
Medical & pharmaceutical bottles

🚀 Quick Start

python

from diffusers import StableDiffusionXLPipeline
import torch

# Load base model
pipe = StableDiffusionXLPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    torch_dtype=torch.float16
).to("cuda")

# Load LoRA weights
pipe.load_lora_weights("shivam909067/sdxl-bottle-marketing-lora")

# Generate image
prompt = "amber glass bottle with dropper, essential oil packaging, marble surface, professional studio lighting"
image = pipe(
    prompt,
    num_inference_steps=32,
    guidance_scale=5.5,
    cross_attention_kwargs={"scale": 0.7}
).images[0]

image.save("bottle.png")

⚙️ Recommended Settings

Parameter	Value	Notes
CFG Scale	4.5 - 6.0	Sweet spot: 5.5
Steps	30 - 35	Optimal: 32
LoRA Weight	0.6 - 0.8	Recommended: 0.7
Resolution	1024×1024	Native training resolution
Sampler	Euler, DPM++ 2M	Any SDXL-compatible

💡 Prompting Tips

✅ Effective Prompts

"frosted glass bottle with silver cap, luxury skincare product, white background, studio lighting"
"emerald green glass bottle, minimalist design, soft shadows, commercial photography"
"matte black cosmetic bottle with pump dispenser, gradient backdrop, professional shot"

📋 Prompt Structure

Bottle description (material, color, shape)
Product context (cosmetic, essential oil, beverage)
Setting (background, surface)
Lighting (studio, natural, dramatic)
Style keywords (professional, commercial, marketing)

❌ Avoid

Complex multi-object scenes
Visible brand names/text (model trained on clean bottles)
Extreme angles or distortions

🎨 Generation Modes

Text-to-Image (Primary)

Generate bottles from text descriptions

image = pipe(prompt, ...)

Image-to-Image

Refine existing bottle photos to marketing quality

python

from diffusers import StableDiffusionXLImg2ImgPipeline

img2img_pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    torch_dtype=torch.float16
).to("cuda")

img2img_pipe.load_lora_weights("shivam909067/sdxl-bottle-marketing-lora")

# Transform rough photo to marketing quality
result = img2img_pipe(
    prompt="professional studio shot, clean lighting",
    image=your_image,
    strength=0.6,  # 0.3 = subtle, 0.8 = heavy transformation
    cross_attention_kwargs={"scale": 0.7}
).images[0]

📊 Training Details

Specification	Value
Base Model	SDXL 1.0
Architecture	LoRA (Low-Rank Adaptation)
Network Dim	128 (upgraded for geometry)
Network Alpha	64 (enhanced edge sharpness)
Training Steps	2,800
Dataset Size	641 images
Resolution	1024×1024
Batch Size	4 (effective, grad accum)
Learning Rate	1e-4
Scheduler	Cosine with warmup
Precision	BF16 mixed precision
Hardware	RTX 5090 24GB

Dataset Composition

Professional product photography
Studio lighting setups
Clean backgrounds (white, gradient, natural)
No visible branding/text
Diverse bottle types and materials

🔧 Hardware Requirements

Inference (Generate Images)

Minimum: 8GB VRAM (RTX 3060 12GB)
Recommended: 12GB+ VRAM (RTX 4070 Ti, 4080)
System RAM: 16GB+

Performance

RTX 4090: ~5 sec/image @ 1024×1024
RTX 4070 Ti: ~10 sec/image
RTX 3060 12GB: ~18 sec/image

📁 Files in This Repo

sdxl-bottle-marketing-lora/
├── bottle_marketing_lora.safetensors  # Main LoRA weights (1.7GB)
├── README.md                          # This file
└── sample_outputs/                    # Example generations
    ├── glass_bottle_01.png
    ├── cosmetic_bottle_02.png
    └── ...

🎯 Use Cases

E-commerce: Product listing images
Marketing: Ad campaigns, social media
Prototyping: Concept visualization for packaging design
Stock Photography: Generate custom bottle images
Design Mockups: Bottle templates for branding

⚖️ License

This model is released under the CreativeML Open RAIL++-M License (same as SDXL base).

✅ Commercial use allowed
✅ Modification and redistribution permitted
⚠️ Must not be used to generate harmful content
⚠️ Must credit base model and LoRA creator

🙏 Credits

Base Model: SDXL 1.0 by Stability AI
Training Framework: sd-scripts by kohya-ss
Dataset: Curated professional bottle photography (Pexels + custom)

📝 Citation

bibtex

@misc{sdxl-bottle-marketing-lora,
  author = {shivam},
  title = {SDXL Bottle Marketing LoRA},
  year = {2026},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/shivam909067/sdxl-bottle-marketing-lora}}
}

Made with ❤️ for product photographers and marketing teams

Report Issues • Request Features • Model Card