Article
9 Oct 2025
Experimenting with AI Video: What We Learned from Veo 3
We’ve been experimenting with Google’s Veo 3, the latest high-quality AI video generator, inspired by the release of OpenAI’s Sora 2 in the US. With access to both, we wanted to see what’s genuinely possible right now, and what’s still out of reach. We’ve produced five short clips, all entirely AI-generated, video, music, and sometimes even voiceover. You can watch them below. The only thing we did was write the prompts. Across around 80 iterations, we learned a lot about what AI can and can’t do, where it impresses, and where it falls apart.
1. Starting Ambitious - The Dance Troupe
We began with something big.
Could we generate a complete dance troupe, moving in sync, ending with a cinematic pan-out and the Yopla logo illuminated in white light behind them?
At first glance, it works - especially on a phone screen. The lighting and rhythm are convincing. But look closer and things start to feel off. The faces don’t quite move naturally. The bodies distort slightly. The text elements (like the Yopla logo) drift or blur.
It’s striking, but not yet reliable. You can feel the potential, and the uncanny valley.
2. Removing People, Adding Energy
Humans are unforgiving critics of other humans - so we took people out entirely.
This next experiment created a starfield and lightning storm, pulsing with colour, ending in the Yopla logo formed from electric light.
This one landed well. It feels immersive, electric, and totally abstract. It’s the kind of thing that could sit behind a conference keynote or at the start of a launch video. The interesting bit? The audio and atmosphere are generated entirely by AI too.
It’s proof that when you strip away realism, AI’s creativity feels more natural.
3. Back to Movement - Embracing Imperfection
Next up, we returned to our dance idea, but leaned into the abstraction.
If AI struggles with limbs and motion, why not use that?
We prompted Veo to create dancers whose movements left trails of light - transforming awkwardness into something almost superhuman.
It’s chaotic and oddly mesmerising, the visual equivalent of turning a glitch into a groove.
4. Simplicity Wins - Penguins, Ice, and Aurora
Lastly, we removed humans again and replaced them with something easier for AI to get right: animals.
A penguin and a fox sit under the Aurora Borealis, light flickering across an ice field where the words “Yopla – Everything Is Possible” appear.
It’s magical - until you notice the fox’s body morphing slightly or that “possible” isn’t spelled correctly. Spelling, it seems, still confuses even the smartest models.
Reality Check
Although this is exciting and enjoyable to experiment with, it shows that the real challenge is not producing video, but producing video that is even average. Creating something good, or exceptional, is far harder.
True creativity is a human pursuit because it is for humans and from humans. AI does not imagine or feel. At best it simulates. It cannot understand story, pacing, or emotion. It can light a scene, but not decide why that scene matters.
The creative process depends on context, emotion, and intent. You need to know what you want to say, why it matters, and how you want people to feel.
This is not just about video. It reflects the wider challenge of technology itself. Understanding how to use it tests our ability to think clearly about problems, outcomes, and value.
We must not outsource our thinking. Instead, we need to recognise where technology is genuinely useful and where it simply adds noise.
There is also an environmental reality. Every generation uses significant computing power. At Yopla, we run renewable infrastructure and offset our usage. Even so, it is worth remembering that these tools are not free. They move the energy cost from the film set to the server.
The Takeaway
AI video generation is fascinating and evolving quickly. It is a useful way to explore ideas, test concepts, and imagine what could be possible. It can help people outside creative industries see and share new ideas.
But AI can only show you what something looks like, not why it matters. That remains the role of people, the thinkers, designers, storytellers, and leaders.
For now, this technology offers space for curiosity, testing, and play. Yet the best results will always come when imagination and intelligence work together.
That is what Yopla is all about.
Try It Yourself
Curious to explore these tools?
👉 Mistral
And if you’re thinking about how to put automation like AI into your own workflows or creative process, get in touch with us.
We’d love to help you make progress possible.