Google Veo 3.1 Lite Released: Video Generation Costs Halved — Who Are They Targeting?

Right after OpenAI announced its $122 billion funding round, Google dropped a bombshell — Veo 3.1 Lite, with a primary focus on halving costs.

The timing is hard not to associate with targeting OpenAI’s funding narrative.

But setting aside marketing considerations, Veo 3.1 Lite itself is quite interesting. It represents a completely different product philosophy:

Not “I can generate the most realistic video,” but “I can generate good enough video at half the cost.”

This difference matters.

The current landscape of the AI video generation track is: Sora leads in quality, but it’s expensive; Kling and PixVerse have cost-effectiveness advantages; other players either focus on quality or price.

Google chose a third path this time — halving costs while ensuring “good enough” quality.

I tried Veo 3.1 Lite’s demo effects. Honestly, compared to Sora, there are indeed gaps in detail handling and physical consistency. But the question is — for most commercial scenarios, do you really need Sora-level quality?

A practical example:

An e-commerce seller wants to make product display videos. Using Sora to generate 10 seconds might cost a few dollars; using Veo 3.1 Lite might only cost a few cents. The quality gap might be invisible to ordinary users, but the cost gap is 10x.

In this case, the choice is obvious.

I think Google’s strategy is smart. Instead of competing head-to-head with Sora on quality, they’re redefining the competition dimension — from “who’s better” to “who’s more cost-effective.”

This is actually a general rule of AI product competition:

Early on, everyone competes on “can it be done”; mid-stage on “how well is it done”; late-stage on “how cheaply is it done.”

Video generation is transitioning from phase two to phase three.

But there’s a risk: does halving costs mean halving quality?

From Google’s technical blog, Veo 3.1 Lite mainly reduces costs through model compression and inference optimization, not simply cutting model capacity. This means quality loss is controllable in most scenarios.

However, I also noticed a detail: Veo 3.1 Lite’s performance in some complex scenarios (like multi-person interactions, fine object manipulation) is indeed inferior to the full Veo 3. This shows Google is also making clear trade-offs — using scenario limitations in exchange for cost advantages.

For developers, what does this mean?

I think it’s a good thing. More cost options mean more flexible business models. You can choose “quality priority” or “cost priority” based on project needs, rather than being forced to accept a unified pricing system.

Additionally, this puts pressure on other manufacturers. If Google can push costs this low, other players either follow with price cuts or differentiate significantly on quality. Either way, users ultimately benefit.

Of course, I still have to say: don’t just look at price.

Video generation models have many evaluation dimensions — generation speed, controllability, editing flexibility, API stability — all as important as cost. Veo 3.1 Lite has cost advantages, but performance in other areas needs more real-world testing.

I plan to use Veo 3.1 Lite on a few real projects in the coming weeks to see what “costs halved” really means in actual workflows.

Speaking of which, I have a question:

Do you think the future of AI video generation is pursuing extreme quality (even if expensive), or “good enough + cheap enough”?

My tendency is the latter. After all, technology ultimately serves business, and business is always sensitive to costs.