Google has unveiled Veo 2, the latest iteration of its video generation model, promising more realistic videos by understanding the nuances of human movement and real-world physics.

Available through Google Labs’ VideoFX platform on a waitlisted basis, Veo 2 aims to revolutionize the video generation landscape.

In a recent comparison, Veo 2’s capabilities were showcased alongside those of OpenAI’s Sora, another major player in the video generation market.

Veo 2 not only understands cinematography language—allowing users to specify genres, lenses, and cinematic effects for up to 4K resolution—but also shows superior performance in internal tests by Google.

These tests measured ‘overall preference’ and ‘prompt adherence,’ indicating that Veo 2 videos were preferred by evaluators over those generated by Sora and other competitors.

Today, we’re announcing Veo 2: our state-of-the-art video generation model which produces realistic, high-quality clips from text or image prompts. 🎥

We’re also releasing an improved version of our text-to-image model, Imagen 3 – available to use in ImageFX through… pic.twitter.com/h6ejHaMUM4
— Google DeepMind (@GoogleDeepMind) December 16, 2024

Despite its advanced capabilities, Veo 2 is still in its early stages. Videos generated will carry Google’s metadata watermark, SynthID, to denote AI-generated content. Although it occasionally produces minor hallucinations, such as extra fingers, Google asserts that these instances are less frequent with Veo 2 compared to other models.

The original Veo remains accessible on Vertex AI, offering an alternative for those unable to access the newer version. The development of Veo 2 highlights Google’s ongoing effort to lead in a market that has seen increasing participation from big names like James Cameron and Andy Serkis, who are exploring AI-generated video content for filmmaking.

In addition to Veo 2, Google has also upgraded its image model, Imagen 3. Now available through its Gemini chatbot, Imagen 3 produces more realistic and brightly rendered images across a variety of art styles, from photorealism to anime. These improvements in Imagen 3 signify Google’s commitment to enhancing the quality and fidelity of AI-generated imagery, catering to a diverse range of creative needs.

Google’s advancements in AI-driven video and image generation are setting new standards, challenging competitors, and expanding the creative possibilities for filmmakers and content creators worldwide.