It is known that AI Tech is progressing rapidly, and Google's recent announcement of the VEO 3 video generator has once again highlighted this. VEO 3 is a step up in terms of realism, and it's the first time I'm adding audio. Often, the output from VEO 3 is hardly distinct from the actual video.
It's not perfect yet, but Veo 3 is sweeping the internet. The viral clip covers everything from street interviews to incompetent stormtroopers. Drawing soldiers from the Galactic Empire is at least one way to ensure consistency between clips. Of course, they all look the same.
Google itself also offers show reels of impressive video clips, including sea sailors and classic violinists. These videos need to look very difficult to tell you that they are AI, and yet they are not always possible.
But these are the final results. How's the creation process? Pay Google for an AI subscription and you can create your own Veo 3 video. There are also a few ways to get in here.
All of this AI video creations should be in context every time. Technology has questions regarding energy use, copyright infringement, threats to the creative industry, and spreading misinformation.
Create a video in VEO 3
Paying $20 a month for a Google AI Pro plan will earn you three VEO 3 video generations per day with the Gemini app, using the faster, lower quality VEO 3 FAST model. If your Google AI Ultra plan grows at $250, you'll get the “best limit” of full VEO 3 access. Google doesn't quantify this accurately, so there are no hard ceilings and can fluctuate based on demand. Each video is fixed at 8 seconds.
If you're using Flow and Whisk tools for video creation, not Gemini Chatbot, that's a bit different. Earn 1,000 AI credits on your Pro plan per month and 12,500 credits on your Ultra plan. The standard VEO 3 video backs up 100 credits, and the VEO 3 Fast Video costs 20 credits. These tools allow you to upscale the resolution to 1080p (720p if you are using the Gemini app).
Create a video with the Gemini app.
Credit: LifeHacker
According to Google's Josh Woodward, Veo 3 Fast is faster and less demanding in terms of processing (obviously) and sticks to the same 720p resolution as a regular VEO 3. Within the Flow app, there is a label that Veo 3 Fast is the fifth quality of VEO 3.
To get started with creating videos, if you are a Google AI Pro subscriber like me, you will need to access the Gemini app on the web (creating mobile videos is currently limited to Google AI Ultra subscribers). Click on the Model Picker in the top left corner to select it and then select it 2.5 Pro (Preview) Or whatever the latest model is when you read this: you can choose video In the text input box, you are ready to run the prompt.
Previously, I tried to recreate an old Sony TV ad using VEO 2. There, thousands of bouncy balls of colour were thrown onto the streets of San Francisco. The results were not great, so I gave the Veo 3 the same challenge. I've gotten a better video back as shown below. It can almost pass as something shot in real life (the sun passing through the trees is fantastic), but it still ignores most of my quick instructions and isn't as good as Sony ads.
This brings us back to the nature of AI that generates us. This is to mimic what you saw before. Veo 3 is trained in many street interviews Vox Pops, and I don't think there are many ads with bouncy balls climbing the hills. It also emphasizes that you may need many prompts to get what you want, and getting the tools to follow the prompts through my AI video test is an ongoing challenge.
There were only 3 generations of VEOs left, so I fasted Veo 3 by recreating the classic “Welcome to Jurassic Park” scene in Spielberg's film. Again, it's better than the VEO 2 effort, but has problems with quick compliance and too many paleontologists. But the dinosaurs (and the sounds of dinosaurs) are well made.
What do you think so far?
Use flow to create longer films
Google also uses the same AI model that comes with the gemini app, and provides a whisk of animation and flow for longer film projects using the same AI model that supports subscriptions. If you don't want to make your own videos, or if you're a free Google Gemini user, you can see what others have made via Flow TV.
Click on Flow in a web browser New Project To get started. Then use the settings button in the top right of the prompt box to select the model you want to use. You will also see the number of credits the generation uses before rendering.
Create a video with Flow.
Credit: LifeHacker
I decided to splash 100 credits for the right complete VEO 3 clip to win a career that will make the Jurassic Park scene perfectly and make AI films. I added a bit more detail to the prompts and some dialogs, but what came out on the other side was equivalent to the VEO 2 output. You can actually choose between two generations, but you can see them here and here.
Again, there's the usual problem in that AI generators really don't know what they're doing here, or how to build scenes beyond what they've seen in other videos. Our brave adventurers look in the wrong direction when one of them delivers a “Wow… will you see the thing itself” line.
The difference between flow and creating extended video is that you can click Add to the scene Start building long enough with one of these generated videos that consists of 8-second chunks. The scene can then be expanded and placed as needed, and the same text and environment is passed from one clip to the next.
My attempt to get John Hammond to enter the scene didn't really work. The original characters remained in the right place enough, but our new characters appeared out of nowhere, and all the sounds were cut out as Flow somehow returned to Veo 2. Especially with only 1,000 units a month, it's clear that I can't switch from high-tech journalism to AI film directors anytime soon.
VEO 3 is still in its early stages, and Google has posted “experimental” labels and flow interfaces on top of it. However, at this point, you need to spend a lot of credits and a lot of time in order to get something consistent and realistic. Hours of effort and trial versions may have entered sophisticated AI videos that you can see living in your social media feed.
