December 16, 2024:
DeepMind Launches Advanced Veo 2 Video Model - Google DeepMind unveiled Veo 2, a new AI model for video generation, designed to surpass OpenAI's Sora. Capable of creating 4K, two-minute videos, Veo 2 offers enhanced physics understanding and camera controls, resulting in sharper footage in various styles. It is currently being tested with VideoFX and aims for broader accessibility via the Vertex AI platform.
Despite its promise, Veo 2 still struggles with coherence in complex scenes and character consistency. DeepMind is working with creators to refine the model, incorporates watermarking to prevent misuse, and acknowledges concerns regarding the use of public data for training models.
