What is Google’s Veo 2? How to Access it, Features, Examples

AI has, and seemingly overnight, mastered text-based chat, voice-chat, and to a large extent, image generation as well. Today, video is the next frontier, and one that’s on the verge of being breached thanks to tools like Sora and, more recently, Google Veo 2.

You’ll find plenty of AI-based video generation tools online, and while they can come very close to producing cohesive videos that can pass for human-made ones, it’s only OpenAI’s Sora and Google’s Veo 2 that have shown any kind of real promise.

What is Google Veo 2?

Google Veo 2 is a project by Google DeepMind, a special-purpose AI research facility working on developing general purpose AI. Veo 2, their latest project, is on the surface a text-video creation tool that takes text prompts and generates video from it. However, Google states that Veo 2 is different from other such tools as they’ve designed it with an implicit understanding of how physics works.

ALSO READ: What is Apple Intelligence, Apple’s very own take on AI?

This, Google says, “Significantly improves over other AI video models in terms of detail, realism, and artifact reduction.” Basically, they’re claiming that video generated from Veo 2 is more realistic.

How to access Google Veo 2

You can access Google Veo 2 by heading to Google Labs’ VideoFX page and joining a waitlist. Veo 2 is still in an experimental phase and hasn’t been made available to the public. Users in select countries have limited access to the model, and unfortunately, India isn’t yet on that list.

Google Veo 2 features

Realistic physics aside, Google Veo 2 apparently has ‘advanced motion capabilities’ i.e. it can create more complex paths for virtual cameras for more creative video. Google says that the user has much more creative control over the physical aspects of the virtual camera, including angles, shot styles, movements, and a complex combination of all the above. 

ALSO READ: Unveiling Android System Intelligence: Insights and applications

Google does point out that complex motion through complex scenes is still a challenge for Veo 2. For some stunning examples of what Veo 2 can do even in this early, experimental stage, head to the Google Deepmind Youtube channel.

Google Veo 2 vs OpenAI Sora

While OpenAI’s Sora — from the creators of ChatGPT — is also very capable and easily the most capable text-to-video tool currently available, it apparently doesn’t have an understanding of the physical world. That said, Sora is available right now (for a fee), and is generating incredibly realistic videos anyway.

As mentioned earlier, Veo 2 is still in an experimental phase, and all we’ve seen so far can be best described as tech demos. Google has, most likely, shown off only cherry-picked examples of Veo 2’s capabilities. While these examples are no doubt stunning, we’ll only truly be able to assess Veo 2’s capabilities when it’s available to everyone.

ALSO READ: Tad AI can use artificial intelligence to compose original songs for free

Realistically, Sora could also improve significantly by the time Veo 2 is officially available, so we can’t really know who the winner is based on current data.