Google launches Veo 2, the latest version of its advanced AI model to rival Pika 2 & Sora

Google DeepMind has introduced Veo 2, the lastest version of its advanced AI video generation model designed to compete with recent releases like Pika 2 and OpenAI’s Sora. This new version significantly enhances visual realism and the depiction of physics, improving the accuracy of movement and human expressions. Veo 2 is integrated into platforms such as VideoFX and Google Whisk, a new experimental tool for visualizing ideas through AI, and is accessible to developers and enterprises via Google Cloud Platform.

A key feature of Veo 2 is its improved understanding of real-world physics, a common challenge in AI video models. It also recognizes different camera types and cinematographic elements, allowing users to specify terms like "18mm lens" or "shallow depth of field" for tailored video outputs. The model supports video generation up to one minute in 4K resolution, reflecting its sophisticated training. Veo 2 addresses common AI issues, such as generating extra fingers or unwanted objects, enhancing output quality.

Future plans include integrating Veo 2 with YouTube Shorts for AI-driven content creation. Access to Veo 2 in VideoFX currently requires joining a waitlist.