Google Cloud published a blog at the beginning of December revealing it had two new AI models – one creates videos, the other more intelligent photos.
Veo is the video generation model and can create short videos from a simple text prompt or an image – real-life or AI-generated.
There's been a lot of talk about artificial intelligence this year, and while it remains in its infancy in terms of capabilities and accuracy, Google has just announced some new tricks for Google Cloud customers that look pretty special.
In a Google Cloud blog post at the beginning of December, the company's senior director of product management, Warren Barkley, revealed that it was investing in AI technology with new models – Veo and Imagen 3. Veo is a video generation model, while Imagen 3 creates more advanced images.
Veo will enable users to create a video from simple text or image prompts. It can do this in what Google describes as "high-quality, high-definition videos" and "in a wide range of cinematic and visual styles with exceptional speed".
Google offered some examples in the blog post as to how Veo is capable of making clips from either existing images or AI-generated images. The first two examples use images created by Imagen 3 to create short video clips, or a man looking up from a book and smiling. There's also a dog standing up and wagging its tail.
The second two examples have created short videos from real world images with the first showing marshmallows roasting over a crackling fire, and the second showing a crowd at a concert.
Google's says Veo has "an advanced understanding of natural language and visual semantics", which allows it to generate accurate video. It explained that the new system can create realistic, coherent movement, so "people, animals, and objects move realistically throughout shots".
When it comes to text prompts, Google offered some examples of how the Veo model would deal with that too, with one example showing a guitar-playing teddy bear (sure), while the other was a sailing boat navigating choppy water.
When it comes to Imagen 3, Google boasted that it represents the company's "highest quality text-to-image model" to date. It can generate much more detail and photo-realistic images, with reduced imaging artefacts.
Both Veo and Imagen 3 will be available on Google's Vertex AI platform. Veo is available as a private preview, while Imagen 3 should be available to Vertex AI customers from this week.
The target is aimed more at larger companies for now, but those include smartphone manufacturers like Honor, who offer models like Imagen on their devices.
