Eli Collins, Google DeepMind’s vice president of product management, first demonstrated the generated AI video tool to the company’s board of directors in 2022. He says this despite the model’s slow speed, high operating costs and sometimes crazy output. It was an eye-opening moment for them to see fresh video clips generated from random prompts.
Now, just a few years later, Google has announced plans for a YouTube in-app tool that will allow anyone to use its Veo model to generate AI video clips and post them directly as part of a YouTube Short. did. “As we look to 2025, we plan to enable users to create standalone video clips and shorts,” said Sara Ali, senior director of product management at YouTube. “You’ll be able to generate a six-second video from an open text prompt,” Ali said. He said it could be helpful. She is adamant that Veo AI tools do not replace creativity, but enhance it.
This isn’t the first time Google has introduced generation tools for YouTube, but this announcement marks the company’s most extensive AI video integration to date. Over the summer, Google released an experimental tool called Dream Screen for generating AI backgrounds for videos. Google plans to update its AI green screen tool with Veo models in the coming months, ahead of a full rollout of generated clips next year.
The sprawling tech company has shown off multiple AI video models in recent years, including Imagen and Lumiere, but it’s looking to coalesce around a more unified vision with the Veo model. “By the way, Veo will be our model from now on,” Collins says. “You shouldn’t expect five more models from us.” Sure, Google may eventually release another video model, but he expects the focus to be on Veo for the foreseeable future. I’m doing it.
Google faces competition from several startups developing their own text-to-video conversion tools. OpenAI’s Sora is the best-known competitor, but its AI video model, announced in early 2024, is not yet publicly available and is reserved for a small number of testers. In terms of widely available tools, AI startup Runway has released multiple versions of its video software, including a recent tool for adapting original videos to alternate reality versions of clips.
YouTube’s announcement comes as generative AI tools become more controversial among creators. Creators sometimes see the current wave of AI as stealing from their work and undermining the creative process. Ali does not believe that generative AI tools will come between the authenticity of the relationship between creators and viewers. “This is really about the audience and what they’re interested in, and not necessarily about the tools,” she says. “But if the viewer is interested in how it was made, that will be made clear through the description.” Google uses SynthID to watermark all AI videos generated for YouTube Shorts. I am planning to include. It embeds an imperceptible tag to identify the video as synthetic, as well as a disclaimer in the description that says “Made with AI.”
Hustle culture influencers are already looking to leverage algorithms using multiple third-party tools to automate the creative process and increase revenue with minimal effort. Will next year’s Veo integration lead to another avalanche of low-quality, spammy YouTube Shorts sweeping through user feeds? “Our experience recommending the right content to the right audience… , I think it works in this large-scale AI world because we’ve done it at such a massive scale,” Ali said. She also pointed out that YouTube’s standard guidelines still apply no matter what tools were used to create the video.
AI art often has a unique aesthetic, which may be a concern for video creators who value individuality and want their content to be unique. Collins hopes that Google’s thumbprint won’t be left across the AI video output. “We don’t want people to look at this and say, ‘Oh, this is a DeepMind model,'” he says. The core goal is to encourage creators to generate AI outputs tailored to what they envision, and avoiding Veo’s overt aesthetics is important to achieving broad adaptability.
“A big part of the journey is actually building something that’s useful to people and is scalable and deployable,” Collins says. “This isn’t just a demo; it’s being used in real products.” He believes that incorporating generative AI tools within the YouTube app will be transformative not only for DeepMind but also for creators. “We’ve never really built a product for creators,” he says. “And we’ve certainly never done it on this scale.”
