Despite announcing Veo later, Google launched this generative video AI model ahead of OpenAI’s rival, Sora.
Google has just released Veo, its latest generative video AI model, allowing businesses to integrate it into their content creation processes. Although first announced in May, three months after OpenAI introduced Sora, Veo has surpassed its competitor by launching first.
Through a preview on Google’s Vertex AI platform, Veo can create high-quality videos with 1080p resolution in various cinematic styles based on text or image prompts.
The videos generated at the initial announcement could be over a minute long. However, Google has not provided specific information about the maximum video length in this trial version.
Some sample videos from this version are on par with what Veo has showcased before. A video based on the prompt “a dog stands up and wags its tail” was described by The Verge as astonishing. From colors and textures to details like fur and the dog’s expressions, everything was harmonious and consistent during the motion.
However, upon closer inspection, oddities can still be noticed. A video of a crowd dancing at a music event becomes unusual when observing the light hitting a hand in the screen’s lower-left corner, or a marshmallow roasting over a fire that remains white.
The latest version of Google’s Imagen 3, a tool for generating images from text, will be available to all Google Cloud customers via Vertex starting next week. Previously, users could only access the tool through Google AI Test Kitchen since August.
Authorized Google users will gain access to new features like photo editing based on prompts and the ability to add elements to images such as brand logos, styles, objects, or unique product characteristics.
Previously, brands like Agoda and Mondelez International had utilized AI technologies such as Veo, Gemini, and Imagen to optimize video ad production, significantly reducing production time.
Google stated that Veo and Imagen 3 incorporate safeguards to prevent the creation of harmful or copyright-infringing content, though these measures are not difficult to bypass, according to The Verge. Additionally, all outputs from these tools will carry copyright seals from DeepMind’s SynthID technology to mitigate misinformation and clearly identify content origins.
With the launch of Veo, OpenAI has fallen behind its competitor. Meanwhile, the promise of releasing Sora by the end of 2024 is running out of time.
Major companies like Coca-Cola have already used AI-generated content in their advertising campaigns. According to Google, 86% of organizations utilizing generative AI have reported increased revenue. This growing demand makes it harder for businesses to wait for OpenAI’s offerings.
