YouTubers will soon be able to play with a host of new generative artificial intelligence-powered tools for creating content, including the ability to generate six-second YouTube Shorts clips, and backgrounds for their videos, using simple text prompts.
The coming updates were announced by YouTube at its Made on YouTube event in New York City today. They’re being enabled through the integration of Google DeepMind’s Veo video generation model, which is a rival to services such as Runway AI Inc.’s Gen-3 Alpha, Adobe Inc.’s Firefly and OpenAI’s upcoming Sora model.
With Veo, it’s possible to generate six-second video clips at 1080p resolutions, based on a wide range of cinematic themes and styles.
The integration of Veo will augment an older AI feature known as “Dream Screen” that was introduced in 2023, giving content creators the ability to quickly generate backgrounds for their videos. Now, users can enter a text prompt into Dream Screen and it will generate four image outputs, which can then be animated using Veo.
YouTubers AI-generated content will be watermarked as such using Google LLC’s SynthID technology, so viewers should have a way to tell between “real” and synthetic content, though there will inevitably be efforts by some creators to circumvent this.
In a keynote presentation at Made on YouTube, YouTube Chief Executive Neal Mohan said he hopes creators will be able to create many more Shorts videos using the company’s AI tools.
“Everything that we showed with AI was meant to really enhance the work that you do, make it faster, more efficient, to bring your creative ideas to life faster,” he said.
The animated video backgrounds in Dream Screen are expected to launch later this year, while the six-second video clips will launch in 2025.
The Veo integration with Shorts was just one of a host of announcements at the event, with other updates in the YouTube Studio app designed to help creators generate titles, thumbnails and video ideas with generative AI. All of these features will become available later this year.
It’s no secret that some social media creators have enthusiastically embraced AI tools to enhance content creation, either by inserting short AI clips, or else creating entire videos using the technology. However, some creators have also raised concerns about the explosion of AI-generated content, and there are fears that their YouTube content is being used to train even more powerful AI models.
Thomas Simons, a comedian with more than 15 million subscribers on YouTube, said in one video that he is far from enthusiastic about the prospects of more AI-generated content.
“I don’t know how I feel about all this AI stuff,” he said. “But it doesn’t fill me with confidence and love.”
There has been criticism that some other social media platforms, such as Facebook, are slowly but surely being overrun by spammy, AI-generated content, while others have raised concerns that AI is helping so-called copycats and potentially violating creators’ intellectual property rights.
YouTube has tried to address the copycat fears. Earlier this month, it announced a new “synthetic-singing identification” tool that creators and artists can use to automatically detect any content on YouTube that simulates their singing voices with AI. It’s also working on a tool that can identify deepfakes of creators, actors, musicians and even athletes, though it’s not clear when that will be made available to users.
YouTube also made a few non-AI announcements, unveiling a new feature called “Jewels” that viewers will be able to send to livestreamers, similar to the “gifts” on TikTok. YouTube said Vertical Livestreams in the U.S. will be the first users to get access to Jewels. In addition, the company is also expanding its automatic dubbing features to cover French, Italian, Portuguese and Spanish.