ByteDance has unveiled Vidi2, a groundbreaking AI video editor that can process hours of footage to create engaging TikTok videos or movies. This multimodal large language model, featuring 12 billion parameters, is designed for superior video understanding, reportedly outperforming even Gemini 3 Pro. Vidi2 leverages advanced spatiotemporal localization to accurately identify timestamps and objects, enabling it to craft narratives based on simple prompts. It introduces tools like highlight extraction and content-aware editing, enhancing TikTok’s Smart Split feature that transforms long videos into concise, shareable clips. With ByteDance’s vast TikTok user data, Vidi2 poses significant competition to traditional AI firms. As the model is still in research, a demo is forthcoming. ByteDance, known for TikTok, is also exploring AI assistants for smartphones, hinting at future integrations with manufacturers, while navigating potential U.S. operation divestitures amidst ongoing regulatory discussions.
Source link
Share
Read more