When it comes to generative AI dominance, companies are actively looking to establish themselves or take the lead by constantly exploring new capabilities. OpenAI has extensively grown and keeps introducing upgrades and innovations to its models through its cutting-edge technology. However, the company has been promising its users an intelligence model for video creation, namely Sora, that remains absent from the market. Google, on the other hand, is revolutionizing content production with a salient head start by launching Veo in private preview.
Veo has taken the lead in generative AI models by finally hitting the market in a private preview
OpenAI was the first to unveil its generative AI video model, Sora, back in February and even showcased its ability to create 60-second videos from text prompts. Google has now taken the lead by introducing its competition, Veo, in a private preview through Google's Vertex AI platform.
Google showcased its AI video model about three months after OpenAI demonstrated its text-to-video model Sora. During the I/O developer conference, Google claimed that Veo came with extensive capabilities that can help generate high-quality videos in over a minute, and users can use varied visual and cinematic styles. While sparse information is available on whether the length would mirror what was demonstrated previously in the videos, the new clips reveal impressive realistic AI videos. The videos are of high quality, making it extremely difficult to tell apart from real footage unless users look at them with extreme scrutiny.
Google's latest Imagen 3, a text-to-image AI generator, will soon be available to Google Cloud customers, with the extensive rolling out next week. The new features would be available to users listed in the allow list and would give them the option for customization, prompt-based photo editing, or even tweaking logos or elements for the generated images.
Google keeps on highlighting that the Veo and Imagen 3 models both come with built-in safeguards that would help prevent infringements but also block any inappropriate content. However, through testing, it was revealed that it is not impossible to bypass the safeguards. The content generated through the models is said to have an invisible digital watermark that is generated with the help of DeepMind's SynthID technology, making it easier to tell if the content, in fact, is AI-generated.
With Google launching Veo, its video AI model, in a private preview, the pressure is on OpenAI to bring about its Sora model. Users have been long awaiting it, and it was promised near the end of 2024.
Follow Wccftech on Google to get more of our news coverage in your feeds.
