Google's visible watermark for VEO 3 Video is not actually visible

AI Video & Visuals


Last week, Google quietly announced that it would be adding visible watermarks to AI-generated videos created using the new VEO 3 model.

And if you look really closely while you scroll through your social feed, you may be able to see it.

The watermark can be seen in videos released by Google and will promote the launch of VEO 3 in the UK and elsewhere.

Ai generated image of a lake nes monster playing bagpipe with


Credit: Screenshot: Google

Google has announced changes to its X-threads by Josh Woodward, vice president of Google Labs and Google Gemini.

According to a Woodward post, the company has added watermarks to all VEO videos except those generated by users with Google's flow tools using the Google AI Ultra plan. The new watermark adds to the invisible Synthid watermarks that all Google's AI-generated content already built into all of its AI-generated content, as well as SynthID detectors that were recently deployed to early testers but still not widely available.

Josh Woodward, VP of Google Labs and VP of Gemini, says in X-Post that it is a visible watermarking “first step” to make SynthID detectors available to more people.

A few weeks after Google introduced VEO 3 with Google I/O 2025, the new AI video model has attracted a lot of attention on highly realistic videos, especially as it can generate realistic audio and dialogs. Videos posted online aren't just fantastic renderings of animals acting like humans, but there are plenty of them too. VEO 3 is also used to generate more mundane clips, including street-on-the-street interviews, influencer ads, fake news segments, boxed videos, and more.

Masculine light speed

If you look closely, you will find obvious signs of AI like smooth skin and false artifacts in the background. However, if you are passively private, you may not consider reconfirming whether the kangaroo casually holds plane tickets. People who are fooled by AI-generated kangaroos are relatively harmless examples. However, according to AI experts interviewed by Mashable for this story, the broad availability and realism of Veo 3 poses a new level of risk for the spread of misinformation.

In theory, new watermarks should reduce these risks. The only problem is that the visible watermark is not. that appear. Video mashables generated using Veo 3 allow you to see the pale white “Veo” watermark in the lower right corner of the video. Do you see that?

Generated images of men interviewing women on VEO 3 street

The VEO 3 video generated by Mashable includes a new watermark.
Credit: Screenshot: Mashable

How's it now?

VEO 3 watermark close-up generated images of men interviewing women on the street

Google's VEO watermark.
Credit: Screenshot: Mashable

“This little watermark is unlikely to be revealed to most consumers traveling through social media feeds in breakneck clips,” said digital forensics expert Honey Farid. Certainly it took me a few seconds to find it and we were looking for it. Unless users know they're looking for watermarks, they may not see it, especially if they view content on mobile devices.

A Google spokesman told Mashable in an email: “We are committed to developing AI responsibly and have a clear policy to protect our users from using AI tools. Content generated by GoogleAI also has visible watermarks embedded in VEO videos.”

“People are familiar with notable watermarks like Getty Images, which is very small,” said a researcher studying people's ability to detect AI-generated content at Kellogg's School of Management. “The watermark needs to be more pronounced, or the platforms that host the image can include notes next to the image. “Check the watermark to see if the image is being generated,” Kamali said. “As time passed, people learned to look for it.”

However, visible watermarks are not a complete remedy. Both Farid and Kamari say they can easily crop or edit videos with watermarks. Farid, who is also a professor of UC Berkeley Information, said, “There's nothing visible about these little – nothing to see.

However, he said Google's Synthid Invisible Watermark is “very resilient and difficult to remove.” Farid said, “The downside is that the average user can't see this. [SynthID watermark] Without a watermark reader, the goal is to make it easier for consumers to see if this type of watermark is included. ”

topic
Artificial Intelligence Google





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *