I Tried the AI Tool That Created These Viral Videos, and It’s Not As Good As You Might Think
We know that AI technology is advancing at a rapid pace, and Google’s recent unveiling of its Veo 3 video generator underscored this once again. Veo 3 is a step up in realism, and for the first time, it adds sound. In most cases, Veo 3’s output is virtually indistinguishable from real video.
It’s not perfect yet, but Veo 3 is taking the internet by storm, with viral clips covering everything from street interviews to incompetent stormtroopers . The depiction of Galactic Empire soldiers is at least one way to ensure consistency between clips, because of course they all look the same.
Google also provided a demo reel of impressive-looking video clips, including a sailor at sea and a classical violinist . You have to look very, very closely to see that these videos are AI, and even then, it’s not always possible.
But those are the end results. What about the creation process? If you pay Google for an AI subscription, you can create your own Veo 3 videos, and there are a few ways to do that, which I’ll cover here.
All of this AI-generated video footage needs to be put into context every time: there are questions surrounding the technology in terms of energy consumption , copyright infringement , threats to creative industries , and the spread of misinformation , all of which we’ve written about extensively before.
Making Videos with Veo 3
If you pay $20 a month for Google AI Pro, you get three Veo 3 video generations per day in the Gemini app, using the faster, lower-quality Veo 3 Fast model. If you go for Google AI Ultra’s $250-a-month plan , you get the “highest limits” for full access to Veo 3 — Google doesn’t give exact numbers, so there may not be a hard cap, and it may fluctuate based on demand. Each video is a fixed eight seconds long.
If you use the Flow and Whisk video creation tools rather than the Gemini chatbot, things are a little different: You get 1,000 AI credits per month on the Pro plan and 12,500 credits on the Ultra plan. A standard Veo 3 video will cost you 100 credits, while a Veo 3 Fast video will cost you 20 credits, and the resolution can be increased to 1080p in these tools (it’s 720p if you use the Gemini app).
According to Google’s Josh Woodward , the Veo 3 Fast is faster (apparently), less processing-intensive, and sticks to the same 720p resolution as the regular Veo 3. It’s unclear exactly what compromises there are in terms of quality (Google’s team didn’t go into detail when I asked via email), but rumor has it that some textures, lighting, and details don’t look as good. There’s a label inside the Flow app saying that the Veo 3 Fast is a fifth of the quality of the Veo 3.
To start creating videos, if you’re a Google AI Pro subscriber like me, you’ll need to head to the Gemini app on the web (mobile video creation is currently only available to Google AI Ultra subscribers). Click on the model selection in the top left corner, then select the 2.5 Pro (preview) or whatever the latest model is at the time of reading this: You can then select Video in the text input field, and you’re ready to follow some prompts.
I previously used the Veo 2 to try to recreate an old Sony TV commercial featuring thousands of colorful bouncy balls being thrown around the streets of San Francisco. The results weren’t great , so I gave the Veo 3 Fast the same task. As you can see below, I got a better video. It could almost pass for something shot in real life (the sun through the trees is gorgeous), but it still ignored most of my quick instructions and wasn’t nearly as good as the Sony commercial.
This goes back to the nature of generative AI, which is to imitate what it has seen before. I assume that Veo 3 was trained on a lot of street interviews with vox pops, and not a lot of adverts where bouncing balls cascade down hills. It also highlights that it can take a lot of cues to get what you want, and throughout my video AI tests, getting the tools to follow cues is a constant challenge.
With only two generations of Veo 3 left today, I commissioned Veo 3 Fast to recreate the classic “Welcome to Jurassic Park” scene from Spielberg’s film . Again, it’s better than Veo 2’s attempt , but there are issues with fast compliance, and too many paleontologists. The dinosaurs (and dinosaur sounds) are well done.
Using Flow to Create Longer Movies
Google also offers Whisk for animations and Flow for longer film projects, using the same AI models you’ll find in the Gemini app, depending on the plan you subscribe to. If you don’t want to make your own videos or are a free Google Gemini user, you can watch what other people have made through Flow TV .
Once you’ve logged into Flow in your web browser, click New project to get started. You can then start the hints by using the settings button in the top right corner of the hints window to select the model you want to use – you’ll also see how many credits will be used to generate it before you start rendering.
I decided to shell out 100 credits for a full Veo 3 clip to better replicate the Jurassic Park scene in full and kickstart my career in AI-powered filmmaking. I added a bit more detail to the prompt, as well as some dialogue, and the output was about on par with the Veo 2 output. In fact, you can choose between two generations, which you can see here and here .
Again, we have the usual problems, in that the AI generator doesn’t really know what it’s supposed to do here, or how to construct a scene beyond what it’s seen in other videos. Our intrepid adventurers are looking in the wrong direction when one of them utters the phrase “wow… you should see that,” and everything from the dinosaurs to the trees looks generic.
The difference between Flow and creating extended videos is that you can click Add to scene on any of these generated videos and start creating something longer, made up of eight-second chunks. Scenes can then be expanded and reordered as needed, with the same characters and environments carried over from one clip to the next.
My attempts to get John Hammond to come on stage were unsuccessful. The original characters stayed in place well enough, but our new character appeared out of nowhere and all the audio suddenly cut out because Flow somehow switched me back to Veo 2. We also got some weird camera shake halfway through. Clearly, I won’t be switching from tech journalism to directing AI-powered films anytime soon, especially on a salary of only 1,000 credits a month.
Veo 3 is still in its early stages, and Google has labeled it and the Flow interface as “experimental.” For now, though, you’ll have to spend a lot of credits and a lot of time working through the prompts to get anything consistent and realistic. Hours of effort and trial runs likely went into the polished AI videos you see in your social media feeds.