Google’s updated Veo model can make vertical videos from reference images with 4K upscaling (arstechnica.com)

Google has unveiled an exciting update to its Veo video AI, now version 3.1, enhancing its capabilities to generate high-fidelity videos from reference images through a feature called Ingredients to Video. This allows users to input up to three images that serve as templates for character animations, backgrounds, and textures, resulting in videos that closely align with the provided references while minimizing random alterations. The upgrade not only improves consistency in output but also supports higher-resolution upscaling to 4K. The significance of this update lies in its strong focus on mobile-first content creation, catering to the rising demand for vertical videos—specifically in a 9:16 aspect ratio optimized for platforms like Instagram, TikTok, and YouTube Shorts. This aligns Veo with current trends in social media, where engaging short videos are key to user engagement. The ability to generate multiple clips with adjustable settings while maintaining core elements intact further expands the creative possibilities for users, making it a powerful tool for content creators in the AI/ML community.