What if the photorealistic beauty of AI-generated video wasn't just a passive view, but an interactive space you could touch?

This project pushes the boundaries of the web, transforming static AI imagery into a living, breathing, and responsive environment. By blending high-fidelity generative video with lightweight CSS masking, we’ve created a seamless bridge between hyperrealism and user interaction.

Live Demo

Instead of traditional 3D modeling, which is time-consuming and often falls short of "perfect" realism, this project uses a purely generative pipeline. By using Google Veo and Nano Banana, we generated a hyperrealistic bloom-and-fade cycle that responds to user input through precise CSS masking.

• Visual Generation: Nano Banana for the initial hyperrealistic static scene.
• Animation: Google Veo for the photorealistic flower bloom-and-fade sequence.
• Interactivity: Claude Code for the custom interaction logic and CSS masking system.
• Core Web: Vanilla JS and CSS for a zero-dependency, high-performance experience.

1. The Generation: We generate two states—a static "idle" garden and a full bloom animation—using state-of-the-art generative AI.
2. The Interaction: The garden is divided into three triangular zones using CSS clip-path.
3. The Masking: When you hover over a zone, the corresponding segment of the 17MB optimized AI video is revealed and played. Because the first and last frames are perfectly color-matched to the static base image, the transition is invisible to the eye.

This project is a static site. To run it locally:

1. Clone the repository:

   git clone https://github.com/yourusername/interactive-garden.git

2. Navigate to the directory:

   cd interactive-garden

3. Serving the files:
   • If you have Node.js: npx serve .
   • If you have Python: python -m http.server
   • Or simply open index.html in a modern browser.

├── assets/             # Optimized AI visuals (17MB interaction loop)
├── index.html          # Core structure and SEO
├── styles.css          # Layout and CSS masks
├── script.js           # Interaction logic and preloading
├── package.json        # Project metadata
└── LICENSE             # MIT License

An experimental exploration of AI photorealism by Vivian Chi.