The fight between Google and OpenAI is real. The latest movements of both companies confirm that they consider the other their great rival, and that they are willing to do whatever it takes to establish themselves as the leader in the generative AI market. Today we have seen the launch to the general public of the ChatGPT web search engine, which is clearly presented as an alternative to the Google search engine; but if you thought that the big G was going to stay still, you couldn’t be more wrong.
And practically at the same time, Google has launched a hook from which OpenAI will take a while to recover: it has managed create a video generating AI superior to Soraand he has no qualms about showing it. Let us remember that Sora was presented in its day as a true revolution in AI, being capable of generating videos of anything the user could think of.
However, continued delays and leaks have lost enthusiasm for Sora, and the public release has confirmed the worst fears: Sora is not as good as she seemedand suffers from many of the common problems of generative AI, such as ‘hallucinations’ and strange object physics.
It is no coincidence that Google has highlighted precisely these two aspects, along with others, in its presentation of Veo 2, its new video-generating AI. If that name sounds familiar to the reader, it is because not even two weeks have passed since the launch of the first version of Veo, making it clear how quickly Google is advancing in this field.
The first comparisons between I See 2 and Sora what social media users like X are doing, make it clear that we are facing a giant step and that the OpenAI alternative is far behind. To begin with, Veo 2 is capable of generating videos at 4K resolution, something that until now was not possible.
Veo 2 brings important improvements compared to the first version. We have already talked about hallucinations, the effect that occurs when the AI loses track of what it was doing; It usually translates into errors such as too many fingers on the hands or objects that appear suddenly. Videos created with Veo 2 are more realistic, partly because These hallucinations are less frequent. Likewise, I See 2 has a better understanding of real-world physics, which is notable when we ask two objects to react to each other.
While OpenAI boasted about artist involvement (although some have criticized it as a marketing ploy), Google seems to have better understood what they want, with new cinematography options in Veo 2. For example, it is possible to indicate the type of lens (such as 18 mm) that we want to use to ‘record’ the content; Thanks to this, it is possible to create different shots or generate effects such as blurred backgrounds that are constantly used by directors.
In the presentation of Veo 2, Google seems to throw a ‘dart’ at OpenAI and the way in which it has launched Sora, which is publicly available (only in some countries and not in Spain). Instead, Google has been more cautious in launching its AI, something that has been “intentional” according to the company to find potential problems and improve security.
In the same way, I See 2 is only available on a limited basis and by invitation, although Google promises that it has expanded the number of users allowed; In the same way, it has confirmed that all Veo 2 videos have an ‘invisible watermark’ that reveals them as having been created with an AI.
