Next, describe what you want to see. If your photo is of a person, give some stage directions: what should they do and say? If your image shows a location like a busy town square or desert landscape, try asking for an unsuspected element like an alien spaceship or sudden snowstorm.

Video generators work by training algorithms on huge caches of images, videos, text and audio. Their output is often imperfect, with odd elements like extra fingers or duplicate details, based on Washington Post testing. But at other times, the resulting video feels indistinguishable from real footage, making it tough to tell when a clip is AI-generated.

AI videos have already trended online, like a fake clip depicting Jewish celebrities giving the middle finger to Ye (formerly Kanye West) after the singer’s antisemitic remarks in February. On the flip side, when a 2023 political ad showed President Donald Trump struggling to pronounce “anonymous”, Trump, who was campaigning at the time, claimed the clips were AI-generated.

AI-generated images and videos have also sparked conflict between the companies that make the tools and content creators, including newspapers and visual artists, who say the firms are training algorithms on copyrighted material.

Last month, Hollywood powerhouses Disney and Universal became the first major studios to file a copyright infringement lawsuit against AI companies when they sued the start-up Midjourney. And as AI models get better, some creative professionals fear that they could lose their jobs to the tech.

Some people worry that making AI video more accessible will lead to more deepfake nudes, or fake sexually explicit images made without the subject’s consent. Deepfakes of celebrities such as Taylor Swift, politicians including Representative Alexandria Ocasio-Cortez (D-New York), popular streamers and regular teenagers have made the rounds online, as some AI generators let users run the program on their own computers without guardrails.

A Google spokesman said the Gemini update is for laypeople and intended to spark creativity, not replace video jobs. The company has a similar tool for professionals, Flow, with expanded capabilities.

Gemini blocks sexually explicit material from its outputs, according to its policy guidelines.

What does this mean for content creators and video professionals?

Video content has replaced static posting as the best method for businesses, influencers and marketers to build a large audience online, said Brendan Gahan, co-founder and chief executive of the influencer marketing agency Creator Authority.

Producing fresh video, though, can be a strain on creators, and burgeoning AI video tools might help them save time fixing bad shots, swapping out audio or adding eye-catching elements. Ideally, that gives them more bandwidth to focus on what really sets a creator apart, Gahan said: the feeling of intimacy they create with their audiences.

Yes, AI video tools will make it easier for more people to make high-quality videos or become content creators, he said. But that doesn’t necessarily mean professional creatives will get boxed out of jobs.

He compared generative AI to a program like Photoshop – it changes the process, but skilled creatives will treat it like another tool in their arsenal. He encouraged professionals, especially social media creators, to learn how to use AI video as quickly as possible.

“There are all these numbers that get thrown around, like 90% of people consume content, 9% engage with content and 1% produce content. If the barrier lowers, more and more people are going to produce content, and that means more and more competition,” he said. “So if you move fast, you can outpace other people and use it to your advantage to cultivate an audience.”

Should we be worried about fake content?

Lowering the barrier to high-quality video also opens the door to more harmful content, warned Miranda Bogen, director of the AI Governance Lab at the nonprofit Centre for Democracy and Technology. Bad actors will use AI video to amplify hateful stereotypes, run internet scams and spread purposefully false information, she said.

“We need more than vague assurances that developers have stress-tested their systems,” Bogen said. “Unfortunately, there are too few safeguards today to ensure companies are not cutting corners on safety testing as they race to launch cutting-edge tools like this.”