Today’s Mad Men aren’t chatty suits trying to convince CEOs that their brands need to reassure people “You’re OK,” it’s tech companies like Google filling the airwaves with AI slop cheap and stupid. Google see you now an AI video generator called Veo, which can take any image, even AI-generated one, and turn it into mini-movies. The videos look semi-realistic, which is good enough for the advertisers Google is trying to sell its tool to.
Google is the first to put its new model in private preview on its business-centric platform. Vertex AI Platform. Véo joins Generation of meta-films and Runway’s video generation tool, although it also beats OpenAI’s Sora model to a wider launch. Last month, artists leaked a version of Sora on the Internet to protest the company’s use of artists with, as they claim, “minimal compensation” to test the tool.
Veo can generate text from text and/or image prompts. Google combines it with its existing Imagen 3 model for any company wishing to create images specific to its brand. AI video generator can create videos with different styles in mind. The company showed off a few examples, including a cartoon man looking up from his desk and smiling. There’s another video of a teddy bear strumming a guitar without any fingers to pluck the strings.
Google says these images include invisible digital watermarksand the model should prevent users from generating anything that might anger the public if they see it appear in their next cookie ad. Imagen 3 further allows users to edit images using its “Inpaint” generation to create new objects in an existing photo or its “Outpaint” tool to enlarge an image, with AI filling in the missing space. It is similar to Adobe’s Existing Firefly Tools in programs like Photoshop.
In its blog post, the Mountain View tech monolith said it was working with advertisers from Mondelez International. Have you never heard of it? It is a huge international conglomerate that owns brands like Chips Ahoy!, Oreo, Ritz, and Tabisco, to name a few.
That’s a lot of brands, and the companies collectively believe consumers demand fast, cheap ads delivered directly to their eyes. Google wrote that its Veo model will enable “the rapid development of large-scale, ready-to-eat visuals” for its many brands in languages specific to all the countries where it sells its snacks.
Google said it was also working with travel app Agoda. The company claimed it was already working on AI ads, including this weird travel ad that will make you break your mouse trying to click the “skip ad” button.
Like most AI video generators, Veo may be limited in the amount of footage it can generate. This is because the generated ads may appear as a jumble of random videos. AGoda’s head of marketing, Matteo Frigerio, said the company created images of “dream destinations” and animated them with Veo.
The end result of these promotions is more ads of questionable quality. The modern advertising model relies on cheap ads made quickly, so much so that they can fill up every ad slot on your social media feed. Jon Halvorson, vice president of consumer experience and digital commerce at Mondelez, said the company is already using Imagen 3 to produce “hundreds of thousands of personalized assets” to reduce “time to market and costs.” “.
AI video visuals may improve over time, although that won’t necessarily change how samey, dull, and lifeless they will be. A survey from the technology research company YouGov showed that about half of consumers dislike ads using AI-generated images, whether of humans or products. Perhaps advertisers will learn the hard way that you can’t underestimate human-made content, even for ads.
Updated 12/04/24 at 10:26 PM ET: This post has been updated to correct and clarify the purpose of the protest around OpenAI’s Sora.