Google's Veo 2 video model will cost 50 cents per second

Hollywood blockbusters like Avengers: Endgame cost a staggering $32,000 per second to produce.

Google thinks it has a bargain for filmmakers: its flagship video model Veo 2 will cost just 50 cents per second.

Sounds dirt cheap on paper, but the real-world math is a lot more complicated.

The meter is running on Vertex AI

Google quietly updated its Vertex AI pricing page, revealing that generating video with Veo 2 will cost 50 cents per second of output.

That adds up to $30 per minute, $1,800 per hour, and around $2,700 for a hypothetical 90-minute film.

Google DeepMind researcher Jon Barron pointed to the massive cost of Marvel movies to frame the pricing as a steal.

Veo 2 was first unveiled in December 2024, initially limited to enterprise clients before expanding via waitlist. One of its standout claims was generating coherent clips stretching two minutes or longer from text and image prompts.

How it stacks up against Sora

The pricing brings a very different billing philosophy compared to OpenAI’s Sora.

OpenAI keeps Sora behind subscriptions. ChatGPT Plus users paying $20 a month receive 1,000 credits for up to 720p, five-second clips, roughly working out to 24 cents per second.

ChatGPT Pro users shelling out $200 a month get 10,000 credits for 1080p video up to 20 seconds, essentially paying around $2 per second for higher resolution.

Meanwhile, Hollywood remains deeply unsettled by the tech. Oscar-nominated films like The Brutalist and Emilia Pérez sparked controversy this year over AI voice manipulation.

Last year, OpenAI faced severe backlash from early Sora artist testers who felt underpaid, gagged by PR restrictions, and used for art-washing.

Why it matters

For independent creators, a ticking clock of 50 cents per second is intimidating.

AI video generation is still notoriously hit-or-miss. Because prompt failures and distorted limbs require multiple re-rolls, paying by the second means creators will be burning real cash on unusable digital scraps.

For enterprise marketing teams and boutique studios, however, Google’s pay-as-you-go model might actually be more attractive than OpenAI’s $200 monthly commitment. You pay strictly for what you render, with support for longer continuous clips.

We think Google’s move is a clear signal that AI video is stepping out of experimental beta labs and into cloud enterprise infrastructure. It forces the rest of the market to be transparent about raw generation costs, but until controllable consistency improves, only well-funded studios will comfortably keep the meter running.