Google has introduced an update for its AI model Veo 3, which allows creating videos up to eight seconds long from a single static image. Now users can generate video clips with AI-generated audio, including background sounds and even phrases spoken by a character in the video. The feature is already available in “preview offering” mode through Google Cloud and works for all clients and partners in Vertex AI Media Studio in 159 countries.
The new feature can be used to create short videos from images of people or products, particularly for social media or advertising. For example, an influencer can upload their own photo and receive a clip where their AI copy walks the runway in branded clothing. Brands can send an image of a product and receive a video showcasing the product from different angles with accompanying audio.
Veo 3 was presented in May at the Google I/O conference. The model immediately attracted attention due to its combination of video and audio, as well as its ability to reproduce realistic motion physics. Google continues to actively develop this direction, and recently Demis Hassabis from Google DeepMind hinted that Veo 3 could be used to create virtual worlds in video games.
Along with the new features of Veo 3, questions have arisen regarding the sources of the model’s training data, as Hassabis stated that videos from YouTube might have been used for this purpose. Some industry representatives are concerned about the risk of spreading misinformation and violating copyright.
