will-z9/my-dog-around-the-world

🐶 Border Collie Generator (LoRA + Stable Diffusion 3.5)

Generate realistic, DSLR-style photos of a personalized black-and-white Border Collie using a LoRA fine-tuned version of Stable Diffusion 3.5. This model captures the visual identity of a specific dog — including its upright ears, fluffy tail, and expressive eyes — and can generate new images in customizable settings.

📸 Demo

👉 Try the live demo on Hugging Face Spaces

Prompt example: a full-body DSLR photo of <dog1> dog running on a grassy hill at golden hour, upright ears, black and white fluffy fur, expressive black eyes, all four legs visible, natural stride

Output sample:

[image]

✨ Features

• —🧠 LoRA fine-tuning on Stable Diffusion 3.5 Large
• —🐾 Personalized <dog1> token representing a real dog
• —🏞️ Support for custom prompts and negative prompts
• —🎛️ Configurable inference (seed, CFG, steps)
• —⚡ Optimized for fast GPU inference with low memory footprint
• —🎨 Realistic DSLR-style outputs with background variety

🧪 Training Details

• —Base model: `stabilityai/stable-diffusion-3.5-large`
• —20 reference photos of one specific dog
• —1000 training steps using DreamBooth + LoRA
• —Prior preservation with 200 class dog images
• —Batch size: 2, gradient accumulation: 4
• —Validation checkpoints saved every 100 steps

🏗 Tech Stack

📜 License

This model is licensed under the CreativeML Open RAIL-M License. It was fine-tuned on top of `stabilityai/stable-diffusion-3.5-large`.

🙋‍♂️ Author & Contact

Built with too much GPU time and an unhealthy obsession with dog anatomy.

For questions, demo access, or technical deep dives: 📧 your-email@example.com 🐶 Powered by [Your Name]

🐾 Want your own dog model?

If you'd like to generate your own pet or object with similar quality and control, feel free to reach out!