README.md

---
title: Style Sync
emoji: 🐠
colorFrom: pink
colorTo: gray
sdk: gradio
sdk_version: 5.36.2
app_file: app.py
pinned: false
license: mit
short_description: Fuses 2 paintings/images
---



---

## 🎨 Neural Style Transfer — A + B = C

Ever wondered what Van Gogh’s brush would look like on your selfie? Or how a castle would look reimagined in Picasso’s palette? **This app lets you do exactly that.**

Upload your own images or remix iconic paintings using **neural style transfer** — a deep learning technique that fuses the *content* of one image with the *style* of another. Watch `A + B = C` come to life in real time.

---

### 🚀 Features

- **✨ Stylize Your Own Images**  
  Upload any two images: one for content, one for style. The result? A stunning hybrid visual that blends their essence.

- **🎲 Pick Random Pairings**  
  Not sure what to try? Click *"Pick Random & Generate"* to randomly fuse masterpieces from the [Famous Paintings](https://huggingface.co/datasets/heramb04/Famous-paintings) dataset.

- **🧪 Adjustable Optimization**  
  Slide between fast runs and deep stylization with the step slider (100–500 iterations).

- **🖌️ Triptych Output** *(coming soon)*  
  Visualize the fusion in a layout: `Content + Style = Output`.

---

### 🧠 How It Works

Under the hood, this app leverages:

- `torchvision`'s pretrained **VGG-19** model
- **Content and style loss modules** with Gram matrices
- **LBFGS optimizer** for iterative refinement
- Hosted on Hugging Face Spaces using Gradio

No pretraining or model uploads needed — everything runs on-the-fly.

---

### 📁 Dataset

This app uses [Famous Paintings](https://huggingface.co/datasets/heramb04/Famous-paintings) hosted on Hugging Face Datasets — curated for visual diversity and artistic flair.

You can remix any pair of artworks, or contribute your own.

---

### 🛠️ Requirements

This app installs the following Python packages:

```
torch
torchvision
gradio
Pillow
datasets
matplotlib
```

---

### 💡 Credits & Collaboration

Built by [Heramb Joshi]

Feel free to fork, remix, or contribute ideas! Pull requests, dataset expansions, and UI upgrades are welcome.

---