Tycia Generator

An interactive generative art installation created for Lithuanian artist Inga Liksaite's project "Tekstai yra čia" (Texts are here).

Description

This project is a Flash-based application that creates dynamic visual art by combining images and text in real-time. It fetches images from Flickr tagged with "tekstaiyracia" and text content from the Tumblr blog tekstaiyracia.tumblr.com. The application generates mosaic patterns from the images and overlays scrolling text, creating an evolving visual experience.

The artwork explores themes of text, imagery, and their intersection in digital space, reflecting Inga Liksaite's artistic vision.

Features

• Real-time image loading from Flickr API
• Dynamic text generation from Tumblr posts
• Generative bitmap mosaics using sampled image data
• Smooth animations with GreenSock TweenLite
• Fullscreen mode support
• Embedded fonts for consistent typography

Technical Details

• Language: ActionScript 3.0
• Framework: Adobe Flash/Flex
• Libraries:
  • GreenSock TweenLite/TweenMax for animations
  • Adobe Core Libraries (crypto, images, net, utils, etc.)
  • JSON serialization
• APIs: Flickr Public Feeds API, Tumblr API
• Output: SWF file embedded in HTML

Project Structure

├── flash-app/            # Legacy Flash application
│   ├── build/            # Compiled Flash output
│   ├── config/           # Build configuration
│   └── src/              # ActionScript 3.0 source code
├── images/               # Image assets from Flickr archive
├── scripts/              # Image management tools
│   ├── generate-images-json.js  # List images with dimensions
│   ├── classify-clip.js         # CLIP AI classification
│   └── classify-images.js       # MobileNet classification
├── images-data.json      # Generated image metadata with AI tags
├── text-data.json        # Text content from Tumblr
├── package.json          # Node.js dependencies and scripts
└── README.md             # This file

Image Management & Classification

This project includes tools to manage and classify images using AI. The images are stored in the images/ folder and metadata is maintained in images-data.json.

Initial Setup

Install dependencies:

npm install

Listing Images

Generate or update the image metadata JSON file:

npm run generate-images

This command is non-destructive:

• Adds new images found in the images/ folder
• Keeps existing images and their tags
• Removes entries for deleted image files
• Updates dimensions if an image file was replaced

To completely regenerate the file (removing all existing tags):

npm run generate-images:replace

Classifying Images

Classify images into semantic categories using CLIP AI model:

npm run classify-clip

This adds tags to each image in 5 categories:

• people - Images with people
• artwork - Artistic content
• exhibition - Exhibition/gallery photos
• texts - Images with visible text
• object - Objects and things

Each category gets a confidence score (0-100%).

Note: First run downloads the CLIP model (~350MB). Subsequent runs are faster.

Combined Workflow

To add new images and classify them in one command:

npm run update

This runs generate-images (non-destructive) followed by classify-clip.

Alternative: MobileNet Classification

For more specific object classification (1000+ categories like "laptop", "car", etc.):

npm run classify-images

Recommendation: Use classify-clip for semantic categorization.

Using the Data

Load images-data.json in your web application:

// Fetch image data
const images = await fetch('images-data.json').then(r => r.json());

// Filter by category
const exhibitions = images.filter(img =>
  img.tags[0].label === 'exhibition'
);

// Find high-confidence matches
const peopleImages = images.filter(img =>
  img.tags.some(tag => tag.label === 'people' && tag.confidence > 50)
);

// Access image properties
images.forEach(img => {
  console.log(img.path);       // "images/photo.jpg"
  console.log(img.width);      // 1024
  console.log(img.height);     // 768
  console.log(img.tags);       // [{label: "exhibition", confidence: 96}, ...]
});

Building

This project requires the Adobe Flex SDK to compile. To build:

1. Install Adobe Flex SDK (or Apache Flex SDK)
2. Set up your development environment
3. Compile with mxmlc using the font config:

mxmlc -load-config+=config/font-config.xml src/com/liksaite/generator/IngaGen.as -output build/generator.swf

Note: The project embeds system fonts (Helvetica) and requires appropriate font ranges for European characters.

Running

Open build/index.html in a web browser with Flash Player support. The application will:

1. Load images from Flickr tagged "tekstaiyracia"
2. Fetch text posts from the Tumblr blog
3. Generate animated mosaics and scrolling text
4. Click anywhere to enter fullscreen mode

Dependencies

• Adobe Flash Player (legacy - no longer supported)
• Internet connection for API access
• Flickr API access (public feeds)
• Tumblr API access

Author

• Code: Matas Petrikas
• Art: Inga Liksaite
• Poetry: Rolandas Rastauskas

License

© Matas Petrikas

Notes

This is a legacy Flash project. Adobe Flash Player is deprecated and no longer supported by modern browsers. For contemporary use, the project would need to be migrated to modern web technologies (HTML5 Canvas, WebGL, etc.).

The Tumblr blog "tekstaiyracia.tumblr.com" and associated Flickr content may no longer be accessible or active.