Google’s new AI video model Veo 2 will cost 50 cents per second
AI video just hit the part of the curve where pricing starts to matter more than demos.
For the last year, most “look what it can do” launches were basically marketing. Now we’re getting a real unit price. And once you have a unit price, you can do math. Math is where a lot of hype goes to die.
Google quietly posted pricing for Veo 2 - the video-generating model it unveiled in December. No big keynote. Just a number sitting on a cloud pricing page. That number is going to shape which use cases win.
The pricing: 50 cents per second
Google’s pricing page lists Veo 2 at 50 cents per second of video.
Here’s what that means when you scale it up:
| Unit | Cost |
|---|---|
| 1 second | 50 cents |
| 1 minute | $30 |
| 1 hour | $1,800 |
That’s not “cheap.” But it’s also not “Hollywood.” It’s a very specific kind of price: high enough that you’ll think twice about generating garbage, low enough that businesses will still use it when it prints money.
The Marvel comparison (and why it’s a little silly, but useful)
Google DeepMind researcher Jon Barron contrasted Veo 2’s pricing with “Avengers: Endgame,” which had a reported production budget of $356 million - or around $32,000 per second.
The comparison is doing two jobs:
- It makes Veo 2 look absurdly cheap next to a blockbuster
- It subtly reframes AI video as “production,” not “software”
Worth remembering: “Endgame” pricing includes actors, sets, reshoots, marketing gravity, and a hundred other things. Veo 2 is just pixels. Still, $32,000 per second is a clean number that makes 50 cents per second feel like a rounding error.
The catch: you pay for seconds you might not ship
Of course, customers aren’t necessarily going to use every second of Veo-generated video that they pay for.
This is the part nobody puts on the slide:
- You generate 60 seconds
- You keep 8 seconds
- You still paid for 60 seconds
That’s not a Google problem. That’s the nature of generative workflows. The “cost per shipped second” is usually way higher than the “cost per generated second,” especially early on when prompts are messy and iteration is the whole game.
Google’s announcement highlighted Veo 2’s ability to create clips that are two minutes or more. That matters because the longer the clip, the more expensive each iteration gets. At 50 cents per second, a two-minute clip is $60 every time you hit generate.
Another anchor price: Sora at $200 a month
Another price to compare: OpenAI recently made its Sora video generation model available to subscribers paying $200 a month for a ChatGPT Pro subscription.
This is a totally different pricing philosophy:
- Veo 2 is metered - you pay by output seconds
- Sora is bundled - you pay $200 a month and figure it out
Metered pricing is great when you’re doing predictable production. Bundles are great when you’re experimenting. Different users, different incentives.
What this means if you’re actually making ads
If you’re generating “cinematic” footage, 50 cents per second is the new baseline to beat. But most marketing teams don’t need cinematic footage. They need performance creative.
They need:
- 20 variations of the same hook
- 10 different creators saying the same line
- Fresh edits every week
- A pipeline that doesn’t require back-and-forth with humans
Traditional UGC is still roughly ~$200 per video once you pay a creator and eat the coordination cost. If your goal is volume and iteration, the modern move is using AI UGC instead of buying one-off creator videos.
That’s why EzUGC exists: instead of paying $200 to a creator for 1 video, EzUGC lets you generate AI UGC videos for $5 each - with better consistency and zero back-and-forth. If you’re the kind of team that ships 50-200 ads a month, the unit economics start to get obvious fast.
If you want to see what that looks like in practice, try EzUGC. If you’re doing the math on scaling, start at EzUGC pricing.