We ran a blind test of Google Veo 3 vs 3.1. Here’s the unfiltered verdict

We ran a blind test with our head of video and senior animator, to find out what these models actually deliver.

“Veo 3.1 brings richer audio, more narrative control, and enhanced realism that captures true-to-life textures.” – The product team at Google Labs and Google DeepMind.

Do seasoned filmmakers agree?

We ran a blind test with our head of video and senior animator, to find out what these models actually deliver – and which one marketing teams should reach for first.

Veo 3 vs Veo 3.1: key differences at a glance

Keep scrolling to see the videos and their comments.

Feature	Veo 3	Veo 3.1 Preview
Camera movement	Conservative, grounded	Dynamic, cinematic
Physics accuracy	Consistent, believable	Inconsistent, occasional breaks
Composition	Strong framing rules	More experimental
Lip-sync	Good	Excellent
Best for	Hero shots, realism-focused content	High-motion, VFX-style content

Veo 3 generations

#1:

#2:

#3:

Veo 3.1 Preview generations

#4:

#5:

#6:

Sian Evans, Head of Video

Veo 3 is the safer, more consistent option. Veo 3.1 Is more dynamic but less believable.

Blind taste test result: “6, 4 and 3.”

Sian’s reasoning:

“What we’re seeing with Veo 3.1 preview is a bit of a mixed bag, to be honest. Yes, the camera movement is more dynamic and the sync work is genuinely impressive – our clients would love that lip-sync accuracy. But when you’re watching the renders back and someone pinches their finger with a caliper, it pulls you right out of the moment.

“Veo 3 might be the ‘safer’ option, but there’s something to be said for consistent, believable framing. We can work with conservative composition – that’s fixable in post or with better prompting. Physics breaks? That’s a reshoot, and with AI, that means starting from scratch.

“The vibrancy in 3.1 is gorgeous, don’t get me wrong. But I’d trade some of that pop for physics that don’t make our audience second-guess what they’re watching. We need to figure out which tool fits which brief, because right now, they’re solving different problems.”

Sian concludes:

“I think we’re seeing Google make a deliberate pivot with 3.1 and it’s revealing their priorities. They’ve clearly fed it a diet rich in high-motion content, VFX-heavy sequences, probably a lot of commercial and music video work where spectacle trumps realism.

“The problem is you can’t just bolt physics understanding onto a model that’s been trained to prioritise visual drama. It’s baked into the architecture from day one. Veo 3 was likely trained on more grounded, naturalistic footage – documentaries, cinema verité, stuff where physics matter because the camera’s capturing reality.”

Arshad Peters, Senior Animator

Veo 3.1 is exciting but the physics aren’t quite there.

Blind taste test result: “1, 2 and 3.”

Arshad’s reasoning:

“From an animation perspective, 3.1 preview is doing some things that genuinely excite me, but then you get these moments where the physics just… aren’t physics. Weight, momentum, collision – it’s like the model learned what movement looks like but not what movement is. As an animator, that’s frustrating because you can see the potential, but the fundamentals are off.

“Veo 3’s composition is very good. It understands framing rules, negative space, visual hierarchy. 3.1 feels like it’s trying to be more cinematic but hasn’t quite figured out the grammar yet.

“I’m hoping this is a preview thing, that they’re testing the flashy stuff first and the physics engine gets refined before full release. But if we have to choose, I’d probably lean on 3 for hero shots and 3.1 for anything where motion and detail trump physical accuracy.”