Google’s new anything-to-anything AI model is wild

Google has introduced Omni, a new family of generative AI models designed to convert various inputs like text, images, and video into new video content. Available within Google's Flow platform, Omni improves upon the previous Veo model by offering better real-world knowledge and character consistency. While testing revealed that Omni can produce convincingly realistic deepfakes that might fool casual observers, it still struggles with temporal inconsistencies, such as morphing objects and sudden character changes. Furthermore, the model operates on a credit-based system, making high-quality, iterative video production a costly endeavor that remains firmly within the uncanny valley.

(Source：The Verge)