Google's VEO 3 has created a splash with Google I/O 2025, the latest leap in the AI-powered video generation.
Get Veo 3 and Flow as features within the new Google AI Ultra subscription, the best access to Google's most advanced models and premium features.
It is designed to turn simple text prompts into hyper-realistic clips using audio, dialogs and sound effects. This is one of the most advanced tools available to casual creators. And after testing it practically, I can say it brings some really amazing results. But it also has a share of hiccups.
This is what I like about Veo 3. And it was something that made me irritated.
What makes VEO 3 right
Please take a look
We've tested a considerable share of AI video tools, including previous versions of VEO, but this latest release is much more user-friendly when it comes to adding sounds and dialogs.
Realism is really impressive – especially due to the fact that you can generate 8-second clips on a computer in under 2 minutes without a full production crew.
Users on the Internet share clips that are hardly indistinguishable from artificial videos. You can get a glimpse of how fast this technology is moving.
Google Veo 3 Realism just broke the internet yesterday. This is 100% AI10 wild example: 1. Pic.twitter.com/qdxzvhoo3g did not occur street interviewMay 22, 2025
Veo 3 is still struggling
For all its strengths, Veo 3 still has a way to go before it gets seamless. Obviously, it's still in experimental mode, so Google is solving the kinks, but the biggest problem I encountered during testing is:
1. Quick interpretation feels hit or miss
Please take a look
The VEO 3 can sometimes struggle with spatial prompts, like when you've searched for an overhead camera angle, but instead got a slightly tilted side view. It seems to prioritize cinematic talent over strict, rapid accuracy that limits creative control.
2. Audio doesn't always work – and why isn't it clear
By default, VEO runs in VEO 2 mode, but this does not include audio. I only noticed this after a few silent clips and a few excavations. To activate audio and dialogs, you must manually switch to Experience Mode under the Quality tab.
Still, lip syncing is inconsistent and dialogue can drop out completely like a badly dubbed foreign film. Plus, subtitles are almost always wrong or wrong.
3. Complex scenes throw it away
Please take a look
VEO 3 shines with a single subject clip, but it can break apart longer and more complicated scenes. The story gets muddy and the characters' interactions often feel stiff or repetitive. If you're aiming to create a multi-character multi-scene story, calm your expectations.
4. The interface still needs polish
There was a moment when I felt the interface was intuitive or unstable. I experienced an unexpected session timeout that erased the generated video, but no recovery options were found.
Furthermore, when I encouraged the model to add dialogs within the scene, I got something that didn't suit the scenario at all.
If this is a powerful tool, the UX still feels a bit rough around the edges.
5. It raises some big ethical questions
Veo's realism is incredible and a bit unsettling. Especially as the technology becomes more accessible, there is growing concern that surreal, AI-generated videos could blur the line between facts and fiction. It also sparks new debates about authors and originality in creative work.
Promising but expensive
VEO 3 is a big step forward in AI video. Especially for casual users who want high-speed, high-quality results. But the Google AI Ultra package costs $249 a month (there are discounted fees for the first three months), and is a sudden price to pay for a tool that still has some notable bugs.
If you want to experiment with video generation or create basic promotional content, Veo 3 is exciting, but isn't necessary yet. But for professional creators, it's worth looking at carefully. Keep the prompts tight, make your expectations realistic, and be prepared to re-render your fingers.
More from Tom's Guide
I'm back to my laptop
