Google Launches Veo: Its Generative AI Video Model, Leaving OpenAI’s Sora Absent in the Market

The competitive landscape of generative AI is heating up as companies strive to assert their dominance and explore innovative capabilities. OpenAI continues to expand its horizon by rolling out upgrades to its technology, yet its highly anticipated intelligence model for video production, known as Sora, is still not available to users. Meanwhile, Google is making significant strides in content creation with the public unveiling of its generative AI model, Veo, currently in private preview.

Google Takes the Lead with Veo in Private Preview

Although OpenAI was the pioneer in introducing a generative AI video model with Sora back in February, showcasing its ability to generate 60-second videos from textual inputs, Google has now captured the spotlight. The tech giant has formally launched Veo in private preview via its Vertex AI platform, marking a critical advancement in generative video technology.

This release comes just a few months after OpenAI’s demonstration of Sora, which took place during their I/O developer conference. Google has touted Veo’s extensive functionalities, asserting that it can produce high-quality videos exceeding a minute in length. Users are provided with a variety of visual and cinematic styles to choose from, although details regarding the specific lengths of videos generated are still sparse. Initial samples reveal that the newly produced content is strikingly realistic, often indistinguishable from authentic footage unless scrutinized closely.

Google launched its Veo model in private preview

In addition to Veo, Google is set to roll out Imagen 3, an advanced text-to-image AI generator, to Google Cloud clients as part of its upcoming features. This rollout, scheduled for next week, will initially be available to selected users on the allowlist, offering a variety of options for customization, including prompt-based photo editing and the ability to modify logos and other elements in generated images.

Google emphasizes that both Veo and Imagen 3 are equipped with robust built-in safeguards to mitigate content infringement and eliminate inappropriate outputs. Nevertheless, preliminary testing suggests that these safeguards are not entirely foolproof. The generated content contains an invisible digital watermark created with DeepMind’s SynthID technology, facilitating the identification of AI-generated materials.

Following the launch of Veo, OpenAI faces increasing pressure to release Sora, which users have eagerly anticipated since its announcement and which is slated for delivery towards the end of 2024.

Source & Images