Google's New AI Media Generation Models
Google recently showcased its latest advancements in AI media generation at its I/O conference. The most significant reveal is Veo 3, a groundbreaking model capable of generating videos with accompanying audio. This technology allows for the creation of highly realistic clips, such as birdsong synchronized with avian visuals or city streets complete with realistic traffic sounds. Google highlights Veo 3's excellence in real-world physics simulation and accurate lip-syncing.
Expanding AI Capabilities
Currently, Veo 3 is exclusively available to Gemini Ultra subscribers in the US via the Gemini app and to enterprise users on Vertex AI. It also integrates with Flow, a new AI filmmaking tool designed to streamline video creation. Flow combines Veo, Imagen, and Gemini to allow users to describe their desired video output in natural language, leaving the technical aspects to the AI.
While introducing Veo 3, Google hasn't discarded its predecessor. Veo 2 remains accessible, offering users enhanced control within Flow. Users can leverage images as references, manipulate camera angles, adjust aspect ratios, and even add or remove objects from their videos.
Imagen 4 and SynthID Detector
Alongside Veo 3, Google also introduced Imagen 4, its latest image generation model. Imagen 4 boasts exceptional detail, accurately rendering intricate textures like fabrics and animal fur. Its improved typography capabilities and ability to generate high-resolution (up to 2K) images in various aspect ratios make it a powerful tool. Imagen 4 is readily available through the Gemini app, Vertex AI, and Workspace applications such as Docs and Slides.
Addressing the increasing difficulty in identifying AI-generated content, Google launched the SynthID Detector. This portal allows users to upload media for analysis, determining the presence of SynthID, Google's AI watermarking technology. While not foolproof, as not all AI generators employ SynthID, it’s a significant step towards responsible AI usage.
1 Image of AI Media Generation:
Source: Engadget