AI Video: Hype vs. Reality
There’s a lot of noise around AI — some of it exciting, some of it terrifying.
It’s either the robot apocalypse or the dawn of creative freedom, depending on which post you read. But when it comes to AI-generated video content, the conversation often leaps ahead of the tech.
So let’s bring it back to earth.
What Can AI Video Really Do Right Now?
If you’ve seen the headlines, you might think we’re just months away from typing a script and watching a fully formed movie pop out the other side.
The truth?
AI video is powerful — but it’s also unpredictable, glitchy, and very hands-on. You’re not directing a scene; you’re coaxing a sequence.
And that changes everything.
Where AI Video Shines
Right now, the best use cases are:
- Music videos
- Adverts
- Mood reels and trailers
- Stylised or surreal storytelling
- B-roll and inserts
Why? Because these formats don’t rely on character consistency, long-form dialogue, or tight shot continuity. You’re telling stories in moments — not full scenes.
Think of it as MTV in the 90s meets Blade Runner dream sequence.
As way of an illustration my very talented contact **Simona Lianza who can be found on linkedin, sent me the video below. It’s a great example of what can be done. A speculative advert done as a showcase. She used ComfyUI to generate the video. She talks of takes for the interations which had all the issues of the 7 fingered monsters. As you can see from the video, it works great as single scenes or shots, perfect for an advert.**
Simona is from Italy, she and I are on exactly the same page.
Yes — AI will lead to job losses. But it will also create entirely new roles and industries. The real challenge isn’t stopping the change — it’s staying ahead of it.
This isn’t a moral debate. It’s a survival issue.
Clinging to the past or trying to ban AI is a losing strategy. Just ask the film editors who refused to adopt Avid in 1989. By 1991, many were out of work.
I make no apologies for saying this clearly: people need to wake up.
This isn’t just about copyright or how the models were trained.
This is about what happens next.
If you work in media, tech, education, or the creative industries — getting ahead of AI isn’t optional. It’s vital. Simona has done just that and I recommend her to anyone wanting to explore the fantastic opportunities AI gives brands to engage with viewers.
Consistency? Still a Work in Progress
You can create beautiful stills with tools like ComfyUI, and you can start to lock in character looks using embeddings, LoRAs, and reference image workflows.
But once you move into motion? Expect your characters to:
- Morph faces
- Lose costumes
- Drift from realism into visual jazz
This is why AI isn’t yet reliable for long-form narrative or anything that needs traditional continuity. You can suggest direction — but you can’t direct. Not yet.
Here is a video i stoyborded in Sora. It was meant to be the journey of a director from a studio to an OB control room during the Sound Check. It was a set from a singer like Dua Lipa. So in ,y prompt I mentioed dancers, without blowing my own trumpet, I have been writing propmu for quite some time. These were the result of my forst iteration. You can see the issues…
Landmarks? Iconic, but Not Accurate
Need a drone shot of the Marina Bay Sands hotel in Singapore?
The AI might give you two towers instead of three. It’s close, but not quite. And that’s part of a wider issue: AI models avoid replicating copyrighted landmarks, and when they try, they approximate — sometimes brilliantly, sometimes hilariously.
The workaround? Start with real plates and enhance with AI. Or use reference-controlled workflows to nudge the model in the right direction.
This was also generated using Sora — an early test for a sizzle reel we’re building for a pitch in Singapore.
Most of us know the Marina Bay Sands Hotel, right?
Take a look at the video… and spot the problem.
This is what the first wave of AI video looks like.
It’s impressive — but it’s also unreliable. The detail isn’t there yet. The control? Limited.
And that’s the point.
So What’s the Reality?
Here’s where we are right now:
- You can build a cast of characters.
- You can design stylised worlds and dreamlike sequences.
- You can generate high-quality mood visuals and short form content.
- But you can’t yet direct a 90-minute feature with full scene control, emotional nuance, and shot-by-shot continuity.
That doesn’t mean we’re far off. The pace of change is wild. We will get to script-to-screen workflows — but today’s tools require human vision, selection, and editing.
It’s more impressionist painting than cinema.
The Takeaway
If you’re in media, this isn’t a threat — it’s a new layer.
AI isn’t replacing directors, editors, or creatives. It’s giving them weird, powerful new tools to experiment with. Tools that work best when you understand their limits.
For now, it’s about exploring the edges. Not automating the centre.
Thanks for reading — and for doing the work.
Keep making things that matter.
Media Consultant & Format Creator
Trusted by 10,000+ media professionals, freelancers, and creatives.
altmedia by Jonathan Glazier
Want to get this in your inbox via Substack?
Or follow me on LinkedIn
