ByteDance has introduced new language models Doubao-PixelDance and Doubao-Seaweed, focusing on video generation capabilities.

These models, integrated into the Doubao AI portfolio alongside previous models such as the Doubao chatbot, are set to launch in early October.

The Doubao-PixelDance model can generate ten-second video clips, while Doubao-Seaweed can produce longer clips of up to thirty seconds.

Both models utilize advanced algorithms to understand complex movements, creating videos that reflect real-life scenarios and fictional narratives.

The Doubao AI family models have gained popularity since their inception, with usage reportedly increasing tenfold since May, reaching over 1.3 trillion tokens daily by October, compared to 120 billion tokens previously.

Research firm LeadLeo projects the AI video-generation market in China will expand to 9.3 billion yuan (approximately US$1.3 billion) by 2026, a substantial rise from just 8 million yuan in 2021.