Google DeepMind launches next-generation video-generating AI to help reduce deepfake risks

The race in the field of generative artificial intelligence (AI) continues to heat up with the emergence of a strong candidate. Google DeepMind - Google's leading AI research lab - officially introduced Veo 2, the next generation of Veo video-generating AI, on December 16.

According to Google, Veo 2 is capable of creating videos longer than 2 minutes with a resolution of up to 4K (4096 x 2160 pixels).

This is a significant step forward, with four times the resolution and more than six times the duration of what Sora, OpenAI's text-to-video model, can achieve.

At the moment, Google's experimental video creation tool VideoFX, where the Veo 2 is offered exclusively, limits videos to 720p resolution and 8 seconds in length.

Google will expand VideoFX's user access this week, and "we'll continue to make improvements based on user feedback in the coming months," said Eli Collins, vice president of product development at DeepMind.

Like Veo, Veo 2 can create videos based on text information or reference images.

The tool can also model motion, fluid dynamics, and light properties more realistically, thanks to various lenses and cinematic effects.

To mitigate the risk of deepfakes, Deepmind said it has integrated its proprietary watermarking technology SynthID into the frames generated by Veo 2.

This is an important measure to identify the origin of the video and prevent misuse.