Google wants to make sure AI videos made with Veo 3 don’t get mistaken as real
Table of Contents
There has been quite the uproar following plenty of AI videos coming out and causing panic with the power they provide. More so with Google’s recent release of Veo 3, as content creators worry, while others imagine the possibility of finishing abandoned projects using the technology. No matter what the purpose, it might be hard to tell apart what’s real and what’s not.
Well, Google is taking matters into its own hands in trying to stop misinformation and make it more obvious. It was already injecting watermarks into its content that could be detected using SynthID, but that’s still not in public release, but rather in early access. Instead, now, as it expands Veo 3 availability to more Gemini app users, it’s trying to make that watermark visible straight away without the need for using another software.
As an additional step to help identify AI content, Google is adding a visible watermark to all videos created in its software. With a small Veo name in the bottom right, it might just make it easier to spot. However, it does not mean that videos generated by Ultra members in Flow won’t get these watermarks, as that membership is made for filmmakers. So it might still be worth checking SynthID on occasion anyway.
Below are just some of the examples Google used for the announcement video of the expanded Gemini App becoming available in the UK.
Google Veo 3 on Gemini baked potato, Source Google
Google Veo 3 on Gemini Royal Corgis, Source Google
Google Veo 3 on Gemini foxes riding a pie, Source Google
Google Veo 3 on Gemini cat on trampoline, Source Google
Is the watermark a saving grace?
Reading that Ultra members don’t get watermarked, it was slightly concerning that it is a bit too accessible. However, considering the cost of AI Ultra, maybe it’s not so bad, as that version costs $124.99 ($249.99 full price) for 3 months. So it might not be a sustainable option for most people.
That means most of the videos created by your average Joe should have the watermark on them. It can be hard to notice, for the most part, especially if there’s a light background behind the Veo name. But it should cut down on a lot of false positives when most people scroll past on a video feed.
However, in a world of vertical videos, that might just get cut off, or a cropped video might just as easily remove it from view, which means having to process it through SynthID whenever that becomes available to check. It seems like too little to really cover and try to cut down on some of the problems AI videos will cause. Hopefully, SynthID is widely available sooner, rather than later.
SynthID watermarking and identifying software
Veo 3 UK announcement
About the Author