It’s always AI at Google I/O! Today, Google announced new AI media creation engines: Veo, which can produce “high-quality” 1080p videos; and Imagen 3, its latest text-to-image framework. None of them sound particularly revolutionary, but they are a way for Google to continue fighting OpenAI’s Sora video model and Dall-E 3a tool that has practically become synonymous with AI-generated images.
Google claims Veo has “advanced understanding of natural language and visual semantics” to create any video you have in mind. AI-generated videos can last “more than a minute”. Veo is also capable of understanding cinematic and visual techniques such as the concept of timelapse. But really, it should be table stakes for an AI video generation model, right?
To prove that Veo doesn’t intend to steal artist’s work, Google also partnered with Donald Glover and his creative studio Gilga to demonstrate the model’s capabilities. In a very short promotional video, we see Glover and crew using text to create a video of a convertible coming home to Europe and a sailboat gliding through the ocean. According to Google, the Veo is able to better simulate real-world physics than its predecessors, and the way it displays high-definition images has also been improved.
“Everyone is going to be a director and everyone should be a director,” Glover says in the video, earning a full Google salary. “It’s all just stories. The more we can tell each other our stories, the more we’ll understand each other.”
Outside of the morbid curiosity of seeing a machine try to algorithmically recreate the work of human artists, it remains to be seen whether anyone will actually want to watch an AI-generated video. But that doesn’t stop Google or OpenAI from promoting these tools and hoping they’ll be useful (or at least make a bunch of money). Veo will be available to some creators today inside Google’s VideoFX tool, and the company says it will also come to YouTube Shorts and other products. If Veo becomes part of YouTube Shorts, it’s at least one feature that Google can have over TikTok.
As for Imagen 3, Google makes the usual promises: It’s said to be the company’s “highest-quality” text-to-image conversion model, with “incredible levels of detail” for “photorealistic, lifelike images” and fewer artifacts. The real test, of course, will be to see how it handles hints compared to the Dall-E 3. Google says Imagen 3 handles text better than before, and it’s also smarter at handling the details of long queries.
Google is also working with recording artists like Wyclef Jean and Bjorn to test the Music AI Sandbox, a set of tools that can help create songs and beats. We only got a brief glimpse of it, but it made for some interesting demos:
The sun rises and sets. We all die slowly. And AI is getting smarter by the day. It seems like this is a huge opportunity from Google’s latest media creation tools. Of course they are getting better! Google is spending billions to make its artificial intelligence dream come true, to have the next big leap in computing. Will any of this really improve our lives? Will they ever be able to create art with true spirit? Check back at Google I/O every year until AGI really shows up or our civilization collapses.
Stay up-to-date with all the news from Google I/O 2024 here!
