# CAN-20 Data Format (DF) Protocol

The initial CAN-20 protocol supported only text-based training data in a QA format. To unlock richer AI development, the enhanced DF protocol significantly expands compatible data types and modalities.

## Enhanced Text Formats

Text remains the core data type, with additional structured formats:

* Markup: HTML, Markdown
* Programming Languages: Python, JavaScript
* Tables: CSV, Excel

## Multimodal Data

And crucially, new modalities introduce audio, visual and video data:

* Audio: MP3, WAV and other formats
* Images: JPEG, PNG and other formats.
* Video: MP4, AVI

## Accommodating Expansion

These exponential gains in potential training data diversity and dimensionality do present storage scalability challenges. We are actively developing L2 solutions to overcome blockchain limitations and fully realize CAN-20’s capabilities.
