Vitalik Buterin gives his approval to TiTok AI for storing images on the blockchain

Vitalik Buterin, one of the co-founders of Ethereum, has given his approval to the innovative Token for Image Tokenizer (TiTok) compression method for its possible application on the blockchain.
TiTok, not to be confused with the popular social media platform TikTok, is a new compression technique that significantly reduces image sizes, making them more suitable for storage on the blockchain.
Speaking on the decentralized social media platform Farcaster, Buterin emphasized TiTok’s potential in blockchain technology, stating that “320 bits is basically a hash. Small enough to go on chain for every user.”
This development could have a significant impact on the storage of digital images such as profile pictures (PFPs) and non-fungible tokens (NFTs).
Developed by ByteDance and researchers from the Technical University Munich, TiTok allows images to be compressed into 32 small data pieces (bits) without compromising quality.
Through advanced artificial intelligence (AI) image compression, TiTok can compress a 256×256 pixel image into “32 discrete tokens,” as outlined in the research paper on TiTok.
TiTok’s 1-dimensional (1D) image tokenization framework breaks free from the constraints of 2D methods, resulting in more flexible and compact images.
Utilizing machine learning and advanced AI, TiTok employs transformer-based models to convert images into tokenized representations by identifying and utilizing redundant information in different regions of the image to reduce data size.
The proposed method aims to represent images with 8 to 64 times fewer tokens than traditional 2D methods, with the research team hoping to provide insight into more efficient image representation.
Despite the similarity in names, TikTok, the social media platform, did not receive Buterin’s endorsement. By highlighting TiTok’s blockchain potential, Buterin adds credibility to this new AI-driven compression method for images.
The team behind TiTok is optimistic that their research will lead to more efficient image representation methods.