Google’s Imagen 2 AI Model Gets Upgraded With Video Generation Capabilities

Google has upgraded its native artificial intelligence (AI)-powered picture era mannequin Imagen 2 with two new capabilities. These capabilities had been introduced at its annual Cloud Subsequent convention in Las Vegas. The AI software is an enterprise-focused product accessible inside the Vertex AI developer platform and will help create emblem components and different visible sources for firms. Now, the tech large has additionally added video-generation capabilities to the AI mannequin and it may well generate as much as 4 seconds lengthy movies.

The video era characteristic is being known as ‘Textual content-to-live picture’ by the corporate. With this functionality, Imagen 2 can create four-second-long movies from textual content prompts. These movies could have a spread of digicam angles and movement. In line with a report by VentureBeat, the generated movies could have 24 frames per second and be delivered at a decision of 360×640 pixels. Google reportedly has plans to enhance these metrics sooner or later.

Google additionally posted a video on YouTube to showcase the capabilities of its AI mannequin. Google Cloud CEO Thomas Kurian reportedly stated in a press launch that animated photos can enhance consumer engagement. The era high quality seems to be much like Runway AI and Pika 1.0 — each can be found for most of the people in addition to enterprises.

There may be additionally a query of deepfakes. That is related since Gemini AI just lately got here beneath fireplace after it confirmed “inaccuracies in some historic picture era depictions”. Shortly after, Google eliminated the picture era characteristic from Gemini. However the firm says it is going to be totally different with Imagen. The corporate’s spokesperson told TechCrunch, “The Imagen 2 mannequin in Vertex AI has not skilled the identical points because the Gemini app. We proceed to check extensively and have interaction with our clients.” As for added safety, photos and movies generated by the AI mannequin will likely be labelled by Google DeepMind’s SynthID expertise.

Aside from video era, Imagen 2 has additionally obtained inpainting and outpainting capabilities. It will permit customers to edit a sure a part of the picture as an alternative of regenerating your complete factor with a brand new immediate. Making granular modifications to a picture is a better technique to get to the specified outcome. The characteristic can also be supplied by Microsoft’s Copilot and OpenAI’s DALL-E 3.