🌎 Ethereum Co-Founder Vitalik Buterin Endorses TiTok AI for NFTs

According to Ethereum (ETH) co-founder Vitalik Buterin, the new image compression method Token for Image Tokenizer (TiTok AI) can encode images to a size large enough to add them on-chain.

On his Warpcast social media account, Buterin described the image compression method as a new way to “encode a profile picture.” He mentioned that if it can compress an image to 320 bits, essentially a hash, it would render the pictures small enough to be stored on-chain for every user.

Buterin’s interest in TiTok AI was sparked by a post from a researcher at the artificial intelligence (AI) image generator platform Leonardo AI. The researcher, identified as @Ethan_smith_20, explained how the method could help encode complex visuals into 32 tokens, aiding in the reinterpretation of high-frequency details within images.

Buterin’s perspective suggests that this method could simplify the creation of profile pictures and non-fungible tokens (NFTs) for developers and creators.

TiTok AI, developed collaboratively by ByteDance (TikTok’s parent company) and the University of Munich, is an innovative one-dimensional tokenization framework, differing significantly from the prevailing two-dimensional methods.

Research on TiTok AI indicates that it can compress 256 by 256-pixel images into “32 distinct tokens.” Previous image tokenization methods, like VQGAN, faced challenges due to their reliance on “2D latent grids with fixed downsampling factors,” which struggled with image redundancies and similarities in close regions.

TiTok AI addresses these issues by using technologies that tokenize images into 1D latent sequences, providing a “compact latent representation” and removing region redundancy. This method could streamline image storage on blockchain platforms while enhancing processing speeds significantly.

Additionally, TiTok AI boasts speeds up to 410 times faster than current technologies, marking a significant improvement in computational efficiency.