Of course, the big conversation surrounding AI video generation is deepfake and misinformation. Google's VEO 3 model can generate surreal videos to make many of us think it's real. This poses serious risks. Was the video you just saw actually recorded by someone? Or was it completely computer generated from a simple prompt?
However, not all AI-generated video output is as sophisticated as it is. Some people are directed towards simple novelty, like Google's photo to video converter. The company first deployed such a tool to pay for subscribers in April. The VEO 2-equipped tool animates images based on prompts, turning the photos into short 720p videos up to 8 seconds. Unlike viral videos created using the company's latest model, the VEO 3, these videos look like party tricks.
Photo-to-Video is deployed in two new apps
Google is working on bringing this feature to more apps, and as a result, it is working on more users. This month, the ability to photograph-to-video came into being for paid users using Gemini. This time we have a Veo 3 (despite the more advanced models, these videos are still short, at 720p.) Perhaps the biggest change here is that we can add audio to these videos. Now, according to The Verge, this feature is heading towards two Google products. Without paying for Google Photos and YouTube shorts.
These new features are once again equipped with VEO 2, so you can get the previous generation video models here. (VEO 3 is clearly coming to YouTube shorts later this summer.) Plus, you won't be able to add custom prompts to the tools. Instead, you can choose from a pre-selected prompt or choose the “I feel lucky” option. That may mean that photos of the lights of the crosswalk will turn into a walk sign dance video. (This is a real example from Google.
Google Photos features generate videos up to 6 seconds long, but YouTube shorts allow you to determine the amount of videos. If you're not satisfied with your generation, you can try again or press your thumb and press the icon. This will help you train your model. (There is also a thumbs up icon here too.)
Like all AI-generated results, Google's photo to video tool warns that it is experimental and “the results may be unexpected.” It's likely a Google cover, so if the animation is doing something offensive or offensive, it's not the company's fault. It's AI: What are you going to do? To that point, there is no confusion as to whether these video outputs were created with AI, as Google says each one is generated with visible labels to indicate its artificial origin. The video also includes an Invisible SynthID watermark. Google bakes all AI-generated products that help you identify actual content from AI content.
The new tool is currently available on US Google Photos for iOS and Android, and will be rolled out next week in YouTube shorts in the US, Canada, Australia and New Zealand.
What do you think so far?
Who are these features?
I admit I don't use many (or any) AI tools in my life, so perhaps I am not the best person to judge the merits of such things. But I wonder: Who is this for?
You can imagine the photo-to-video tool to be interesting for generations or two. Look! We have animated this image of a family dog. And now the doodles I drew this morning are moving! And now I'm bored.
There is something novelty about these tools, but in my opinion it wears out quickly. It may be fun to share these AI videos with friends from time to time, but I would rather see real photos than fake animations derived from those photos. If I try to argue this feature beyond novelty, it falls apart for me. Google uses an example of a photo of a smiling man. Veo2 animates a man throwing leaves into the air. The animation is not only irrelevant to what is happening in the image in the first place, but it is just a false memory and never actually happened. Please give me something in the stationary version did It happens anytime.
