Generative AI models, Veo and Imagen 3 have been revealed by Google as it rolls out private access to these models. From Wednesday ie. December 4, users of the vertex AI Google Cloud package can make use of Veo to create videos from text prompts or images. Next week, Google is also going to introduce its latest text-to-image framework, Imagen 3 to the same group of users.
Google claims to be the very first major cloud provider to offer an image-to-video model with the launch of Veo. In contrast, OpenAI’s Sora model is still limited to a specific group of artists, researchers and academics. Although the company has already dropped the hint at 12 days of product demonstrations that started from 5 December; it is expected that this may change soon.
Veo, according to Google produces 1080p video which is consistent and coherent and can extend beyond a minute in length. The tool operates smoothly and generates results from both prompts as well as images allowing users to begin with either AI-generated images or human-created visuals.
Examining the sample footage shared by Google stated that similar to several AI models, Veo can encounter challenges with effect and cause. In a clip that shows marshmallows roasting, the treats fail to brown and char as they are being heated by the campfire, in addition to this, artefacts are quite evident, especially in the hands that are shown in the concert footage.
As for Imagen 3, Google emphasizes that this model creates the most practical and best quality images from simple text prompts and it exceeds the previous Imagen version in detail, artefact reduction as well as in lighting.
Google is eager to encourage more of its enterprise clients to adopt generative AI and per the research, it has done, 86 per cent of companies using generative AI in production have reported an increase in revenue.
