THINKING

Thinking on AI video, advertising and creativity from Republic of Imagination. Founder Julie Seal writes about AI filmmaking, creative direction, craft, production and the changing advertising industry, including her regular Ay? Ey. column for The Drum.

AI VIDEO, CREATIVITY & CRAFT

Adland’s old guard won’t solve AI video’s cowboy builder problem

AI video is booming, but bad execution risks damaging trust in the medium. The answer isn’t handing AI production back to adland’s old guard, but combining new AI-native talent with genuine creative craft. Republic of Imagination founder Julie Seal argues for a new generation of AI video specialists who understand both technology and storytelling.

So, AI video can actually be good... in the right hands

Why AI video isn't inherently brilliant or terrible. It's what happens when you put it in the hands of people with ideas, taste and creative direction that matters. The story behind Republic of Imagination's pivot into AI video production.

A nan-friendly guide to AI video in advertising

AI video without the tech-bro bollocks. A plain-English guide to generative AI video in advertising, from models and tools to consistency, synthetic media and what it actually means for production.

Inside Studio Ru's strange, smart bet on AI video

What actually separates good AI video from slop? A conversation with Studio Ru founder Casey Bird about AI production, creative direction, taste, ethics and why knowing what to make still matters more than knowing which buttons to press.

AI video's biggest creative problem is still taste

Ten things learned from actually making AI video, including blandly beautiful avatars, posh voiceovers, representation, consistency and the occasional accidental demon. Because better AI tools don't automatically make better creative work.

AI video is only safe if you keep your receipts

AI video production comes with some properly grown-up questions about copyright, provenance and commercial use. Why brands and agencies need transparent workflows, clear records and an audit trail of how AI-generated work was actually made.

Omar Karim is waiting for AI filmmaking's 'Gorilla' moment

A conversation with AI filmmaker Omar Karim about moving beyond AI as a production shortcut and towards genuinely AI-native storytelling. The opportunity isn't to recreate filmmaking more cheaply. It's to make creative work that couldn't have existed before.

Exactly when does an AI-generated ad become slop?

“AI slop” gets thrown at almost anything made with generative AI. But bad advertising existed long before AI did. A look at where the line really sits between lazy generation and properly conceived, directed and crafted AI advertising.

The BS, red flags and snake oil in AI adland

As brands rush into generative AI, an entire industry of experts, claims and questionable promises has appeared alongside it. A look at the red flags, bullshit and snake oil worth spotting before you hand over a brief or budget.

Taste: the thing no AI prompt can give you

AI can generate extraordinary images and increasingly extraordinary film. What it still can't give you is judgment. Why taste, cultural understanding, creative direction and knowing when something is actually good become more valuable, not less, in an AI production world.

I think you're all too chicken to admit AI can help

A conversation with filmmaker Nick Tasker about a very good AI-made ad involving a unicycling chicken, and why parts of UK advertising still seem reluctant to embrace AI production. The technology isn't the idea. The question is what talented creative people choose to do with it.

IN THE PRESS

Republic of Imagination: Julie Seal puts craft back at the heart of social with new studio

Little Black Book on the launch of Republic of Imagination: why Julie Seal founded the studio after seven years at Meta, the thinking behind the name, and the original ambition to bring proper creative craft and maker-led production back to social advertising.

BUILDING REPUBLIC OF IMAGINATION

I'm founding a creative agency. Here are my trials and tribulations

The beginning of Republic of Imagination, documented in public. From decades as a creative director and seven years inside Meta to funding, naming, positioning and trying to work out exactly what kind of agency should exist next.

OK, I think we've finally launched our new agency

Republic of Imagination officially enters the world. A founder diary about launching an independent creative business, finding a genuinely differentiated offer and why our obsession with makers, craft and creativity became central to the agency.

It's my birthday and I'm starting to think my new agency will see one too

A few months into building Republic of Imagination, the terrifying bit: it starts to feel like it might actually work. An honest founder diary about clients, momentum, uncertainty and building an independent creative company from scratch.

Laying down the law and getting work under our agency's belt

The less glamorous bits of starting a creative agency: contracts, legal stuff, commercial realities and turning an idea into an actual working business, alongside the much more enjoyable business of getting real work out into the world.

My new agency works in Social 4.0... I'm not saying what that is yet

Republic of Imagination's thinking about what comes after traditional social media marketing: entertainment, algorithms, creators, communities and brands making things people might actually choose to watch.

More from Julie Seal in The Drum

More columns and commentary from Republic of Imagination founder Julie Seal on AI, advertising, creativity, social and the creative industries.

CREATIVITY, CULTURE & THE INDUSTRY

Mandatory office returns exclude parents, disabled people and their amazing ideas

Why blanket return-to-office mandates risk excluding talented people who don't fit a traditional five-days-at-a-desk model, and why flexibility can make the creative industries more diverse, accessible and creatively interesting.

Sure, I'd love to work in an office with 'masculine energy'

A response to the idea that companies need more “masculine energy”, via decades of experience in advertising, Meta and some considerably less conventional workplaces. What good leadership and genuinely diverse creative cultures actually look like.

There are two types of creative director: builders and breakers. Which are you?

Creative direction isn't just judging ideas. It's knowing how to make people and their work better. A case for the creative directors who build confidence, talent and ideas rather than breaking them down.

Ay? Ey. 10

I was reminded what a cruel word ‘redundant’ was when chatting to the loveliest Simon Helm today… like you’re an ‘unnecessary, redundant human’ when it happens. Well it’s happened to us all (me thrice, the first helped me get on the property ladder, the last went miscarriagey, so Ws and Ls, as the young people say). 

But very far from being a redundant human is the quite excellent Simon… former Head of Art and AI at Fold7, now founder, maker and chief AI-problem-fixer at Studio Helmo. We had half an hour. We spent some of it discussing standing desks and quite a lot of it trying to fix my own Seedance workflow glitch. Sorry not sorry. This is the rest…

Julie: Okay so firstly, you had an actual AI role inside an actual ad agency, which weirdly makes you the first person I’ve spoken to for this column who has. What happened?

Simon: I was Head of Art and Head of AI at Fold7. Then I was made redundant just over a year ago. And the word ‘redundant’ is such a horrible thing. It feels like somebody’s saying: you are a redundant human. I felt bruised by it and I knew I needed an identity shift. I didn’t want to freelance and carry on doing exactly the same thing. I’d been Head of AI, but it was frustrating because nobody really wanted to listen.

People could use Midjourney for their visuals and that was all they felt they needed. They didn’t want to progress the knowledge.

Julie: Oh so weird! Was that simply because it was early? Would it be different now?

Simon: Yes and no. Creatives just want to create. They want to come up with the idea. They don’t want to ‘do AI’. They want to go to South Africa and shoot it with Dougal Wilson.

Julie: This is such a hilariously accurate ad-creative stereotype, lol.

Simon: But if you see AI as the lowest common denominator, you’re shooting yourself in the foot. I thought it was fantastic. You could suddenly do all this stuff.

I think the feeling was that craft was at a low point with AI, and people wanted to work with retouchers and 3D artists and directors and illustrators… all the external craftspeople. Creatives were quite purist. Maybe AI was seen as this non-human part and they just didn’t see anything in it.

I did. I found it liberating. I was saying: everyone, you can do this. You can make it move now. There’s so much you can do here.

Julie: This is the bit I love too. Seeing the fruits of your labour instantly. Making as part of thinking instead of sitting around talking about making and then writing a deck about the chat about the making. So how did being made redundant become Helmo?

Simon: Recruitment agents made it quite an easy decision. Nobody returned my calls. Nobody was helpful. Everyone was saying, ‘I think I can get you something for half the money you were just earning.’ It looked so bleak and I felt so depressed by it all. I thought: this isn’t the way.

AI was burgeoning, and while I was bored between calls with recruiters I was making things. I started putting the work on LinkedIn and got loads of really good responses almost immediately. I never got that response from my commercial work.

So I just ran towards the sun. I followed where the positive rays were coming from.

Julie: Lovely. Depressing. But lovely. Run into the light!

Simon: I knew I loved art direction, and this was such a visual space. I thought I could bring quality, craft, creative direction and ideas to it.

I slowly picked up clients. I did some random work involving robots for a fintech brand, then work for the activewear brand Oner. I worked with an agency called Anything is Possible. They said they wanted a Cannes-winning banner for their website, so obviously I made them a musical.

Julie: Obviously.

Simon: ‘Anything is possible’ sounded like an anthem, so I used Suno to write a musical inspired by Grease and La La Land, then created the visuals. They never used it, but they asked me to create their identity afterwards.

I made this whole AI photoshoot of little visual poems about anything being possible. A feather outweighing a kettlebell. A tiny truck carrying a mountain. A flying pig. I animated all of them.

I was using my advertising background, design background and AI together. People liked the film work, they liked the fashion work, and it grew from there.

Julie: Amazing. And now you’re finishing what you’ve called the world’s first AI sitcom?

Simon: Yes. It’s 24 minutes long and it has been epic.

It’s almost entirely dialogue. There’s a tiny bit of action, but it’s mostly comedy timing, which is hard. The script was written by the client during a period of insomnia in 2004. So it isn’t AI-written. I created the characters, environments, voices, attitudes, how they behave and all the interactions. I put the meat on the bare bones.

I was interested because nobody had done it. I’m interested in firsts… the things people might look back on and remember as the first time they saw something.

I’ve spent about £3,000 in Seedance credits on it so far. It was meant to take a month. It’s taken more like three or four.

Julie: Ouch. This is why we’ve productised our prices at Republic of Imagination. Nobody wants to pay for how many times a model has randomly grown an extra arse or decided a Northern man is now Spanish. They want to pay for the final asset.

But the random extra arses are very much our problem.

Simon: Exactly. There’s usually a workaround to any problem in AI, but you have to know all the different ways of fixing it. Sometimes you don’t need to rerun it. You can fix something in the edit, put a face over two clips where they meet or use an over-the-shoulder shot so you don’t have to rely on bad lip sync.

You have to think differently about the whole workflow. Grading lots of different clips later is difficult, so I try to get the image almost graded before I feed it into the film model.

Julie: It’s a completely arse-about-face way of working compared with what we knew, really. I’ve found ideation has changed too. Sometimes I make the early prompts deliberately vague so I can explore through doing.

We didn’t do that in agency land. You came up with the idea first, then got permission and money to make it. Now the machine can be part of finding the idea, which is really rather lovely.

How has running Helmo actually been though? Beyond exciting sitcom adventures and credit-induced financial ruin?

Simon: Nail-biting. It was fine while the redundancy money lasted, then I had a really dead February and March. I’ve got a mortgage and two children and I was approaching so many people. It was a worry.

Then in April things suddenly picked up. The sitcom came in, more leads appeared and I started getting paid by AI companies to try and talk about new tools. I’ve also worked with agencies including AMV BBDO and Interbrand, and I’m talking to some premium fashion agencies and a fashion brand in Italy.

At the moment Helmo is just me. But I’m secretly interviewing people whenever I meet them.

Julie: Ha! Same. What are you looking for?

Simon: Grit. When you work in AI, you need somebody who can find that bit of grit when everything is going wrong.

On the third round of changes, you might think, ‘I don’t see the point in this. Every part of my body doesn’t want to touch it.’ But you still make it good.

I want to find another two of me so the studio can take on more work without losing the quality.

Julie: Honestly quality control is everything to me. Sure we can make shite if you want. But you shouldn’t want. You’re also teaching agencies now. Are people more receptive than when you were trying to do it from the inside?

Simon: Everyone wants to know now. That’s why I love doing it.

The problem is that everyone wants to learn, but nobody really implements it. When I teach, I insist people do it with me so they don’t have to make the leap alone afterwards. But people are busy. They leave it, forget it and don’t use it… like any course you buy.

You need somebody inside an agency who is always on it. If you don’t have that, you simply won’t know what’s there or how to use it.

I know people repeat the line that AI won’t take your job, but somebody who can use AI will. I do think it’s true. It’s going to be like Photoshop. If you can’t use it, there will be a problem.

Julie: I agree, actual use on actual jobs. Failing, fixing, finding workarounds like you do, hybriding, tweaking and spaffing credits.

Simon: Exactly.

We then ran out of time, and so I asked Simon to follow up with some unpopular opinions about AI on email. He did just that, et voila (gold):

1. Some agencies call themselves 'AI integrated' when really what they mean, is that they have a MidJourney & ChatGPT subscription and have played with Generative Fill in Photoshop. They have some tools and some curious employees but they aren't production ready and don't know how to be either. I wonder if they think that's the extent of it or maybe they aren't taking it seriously until they realise it's not going away and now they are behind.

2. LinkedIn is flooded with AI Creative Directors but if you look deeper, most are either unqualified at being a CD or they aren't really THAT 'AI' fluent. Lots of self proclaimed lofty titles out there....

3. Recruiters aren't doing enough to understand what an AI Creator, AI Art Director or Creative Director needs to know or do - they are getting feedback from agencies about what the agency wants, but the agency doesn't always know either. This results in hires that don't deliver - perm or freelance - I've been called in to unpick many an unresolved issue.
Not all AI Art Directors know how to artwork to production standards and know how to make painful but necessary corrections to films and stills.
Just because you can make cool MidJourney 8.2 images, doesn't make you a 'fully' AI Art Director, it makes you more of an Art Director who can also use AI.

4. If you want to be great with AI image and film, you have to be obsessed with it, you have to follow all the latest tips and trick, use cases, new features, new models, model updates, Lives, and webinars... the lot, you need to make this your life or you just won't be able to keep up and you won't be able to label yourself fully AI native.

5. Agencies that sleep on AI and don't create connections with great commercially savvy, production ready AI creators / Art Directors / Makers etc will end up losing out by either paying through nose or not being able to book them, because there aren't that many of us to go around.

That’s the opinion bit, I suppose. Agencies don’t need another inspiring AI talk so everyone can nod thoughtfully and return to exactly the same workflow. They need a Simon. A person who knows what good looks like, knows what might fix it and will still be there on the third round when it all goes tits up.

Preferably with £3,000 of someone else’s Seedance credits.

Ay? Ey. 9

Has Adland Gone Chicken? One of Republic of Imagination’s esteemed collaborators, the delightful Nick Tasker, last week made a spec commercial about a hen riding a unicycle. It was/is, frankly, glorious. The shots were well-chosen and charming. The DOP did some lovely stuff. There were nice little touches in it, simple but slick storytelling, and a smashing pay-off. Nice little score and very satisfying VO too. It was all-round good blummin’ advertising. Well done that team. Except of course it wasn’t a team. It was Nick. In his studio. Alone (well maybe with multiple appearances from his kids as it’s the hols). I don’t really care or clock how it was made, it made me smile, but it seems the industry over here has a bit of beef with this chicken… I’ll let Nick explain…

Nick: Like all my ads, the idea was written the old-fashioned way. On the back of my hand, in biro. But unlike the 473 ideas before it, this one didn't get washed down the plughole. Thanks to AI, and a little help from Alibaba, who funded the production, the idea became a reality.

Then I put it on LinkedIn and my inbox went bananas. Three Hollywood agents. Two New York agencies. Four established production companies. Six podcast hosts. Oh, and one angry meat eater.

But from the London ad agencies, where I spent two decades building relationships and learning my craft shoulder to shoulder with so many greats: crickets.

Nobody was rude, by the way. Just a little sniffy. It's something I've noticed a lot, and I've started calling it the great AI divide. Across the pond they're embracing AI production and quietly planning Super Bowl commercials. Over here we're still shaming people for using ChatGPT in the office.

I swear you could tell someone Guinness ‘Surfer’ was made with AI and you'd get the same "yeah, it's fun" then change of subject. You know the face, the one when someone brings six-quid wine to a posh dinner party and the host forces a smile.

Or there's the predictable "AI slop" comments. I actually agree with that most of the time, but in truth, 98% of all commercials have always been slop. Granted, AI is at about 99% right now, but catching up, fast.

(I do hope everyone who's ever typed those two lazy words has cancelled their Netflix subscription, since they announced around 300 shows are using the tech.)

And the old-schoolers will argue it's about quality. Bollocks. (We've seen your reel, mate.) It's F.E.A.R.

Fear of change. Fear of irrelevance. Fear of having to learn something genuinely new at 42 and a half, with three kids, a dog and a mortgage (been there).

And, if we're being properly honest, fear of not being able to fly business class to Cape Town to shoot a mediocre yogurt ad then spend a month in some fancy VFX studio in Soho sipping lattes, eating bento boxes and telling your partner over a Waitrose fish pie how "this is gonna be the one". (Sorry, love.)

But being sniffy is a luxury, and I'm not sure we can afford it any more.

While we're stroking our collective hipster beards deciding whether AI counts as real work, the Americans are already using it. Not as a lunchtime experiment, but on the biggest, most expensive, and potentially best work in the business.

And it's not as though UK adland is having a vintage decade otherwise. Budgets down, headcounts down, work getting safer, which is to say riskier, every year. And clients running off to in-house, flipping us the bird on the way out.

We've spent ten years being told this industry is in trouble. Then someone turns up with a magic wand that means our only limit now is the ideas we can dream up, and our response is to fold our arms and mutter about craft.

Too many still treat GenAI as another nail in the advertising coffin. But if you put down that latte for one moment, you might realise it's actually a fucking lifeline for adland.

Creatives need to be the ones pushing things, taking risks, thinking bigger and leading the more nervous ones forward. Now you can make the one where a bank CEO fights a badger in a lift. The one where a Yorkshire terrier drives a combine harvester through a wedding. The one where two thousand garden gnomes walk into the sea at dawn (ok maybe not that one, that one is definitely slop).

Come on, Adland, what are we waiting for? Don't be chicken. Let's fly again.

Okay back to me (Julie), thanks Nick… and, well, right? 

Why is it that here in the UK we are often the laggards, the slowest of the slow, the ocean liners turning around? Through and since my time at Meta I’ve seen the same social achingly slow approach to change from UK agencies (not elsewhere). It’s pretty frustrating. (Note: most social agencies are still in 2023 at best, kills me).

Sure, some cultures are a switch, some are a dial… but bloody heck we could do with flicking the switch on this one and accepting the AI fate, it ain’t going anywhere. My fave hilarious quote from someone I know when he faced work struggles and I suggested he get into learning AI was “No, I don’t like it.” (He’s still sadly unemployed). But it made me lol because AI don’t care about your feelings. The industry, all industries, the whole world is tumbling head first into AI and it’s not going to back up just because it ain’t your bag. Smash up the knitting frames if it makes your feel better. Struggle on with candles. Write a proper paper letter to your MP, perhaps. 

But also maybe show proper creativity some love, without British Empire superiority, snobbery and churlish indignance. Nothing good comes from fear.

Ay? Ey. 8

Martin Parr was a snob. Yep, controversial I know, especially as he was beloved by so many. Yes, he also photographed middle class people, but it’s the fun-poking depictions of working class Brits that brought him huge fame. Maybe he just loved people and culture. Maybe he just found human behaviours and their shonky tastes amusing. He was great, but he was also a snob. And that’s okay, the working classes can take it, they don’t take themselves nearly as seriously as the middle and upper classes anyway. As Grayson Perry explores taste in his brilliant documentary series on the subject, where he argues that taste isn’t objective, it’s actually a brilliant marker of social class and identity. 

Taste is a bloody funny thing. And I agree with Perry, innately cultural. European tastes for design are nothing like US tastes for design, for example. We believe our taste, craft, subtly, humour and style to be far superior in advertising and design. You’ve only to watch a few minutes of American commercial TV to wince and whine at the visual and audio assault of craftlessness.

I used to live with a Swiss typographer who bemoaned London’s ‘ugly’ street signage, claiming that such awful, tasteless public design simply wouldn’t have been allowed in Bern. (He also pressure-cookered broccoli though so who was he to know about anything).

I’ve been re-editing our Republic of Imagination showreels, updating the old website and generally giving the ‘new pivot to a AI video studio’ a bit more of a proper identity. That requires me, as always as Creative Director (or CCO), to be the arbiter of taste for it of course. As Creative Directors that ‘what’s good and what’s not good, what works, what communicates, what looks right and doesn’t’, that’s literally the crux of our skillset. It’s built from years of experience, of honing, of failure and success. Of being able to articulate what is or isn’t working and why, and of building ideas collaboratively. The best creative directors are smart, generous, kind, discerning, considered and have sharpened all those skills for years working with delicate egos, nervous clients and other very smart people. 

So that’s why I think that taste is way more important than tech when it comes to being an AI video ‘maker / director / filmmaker / practitioner’. Tech enables, and it’s important to know how to achieve a fab result. But if you don’t know why the result you are trying to achieve is good or bad in the first place, what’s the hope?

‘In-housing’ is popular right now for much of the AI-visual stuff made in adland, but it’s mostly being farmed to junior artworkers, designers and art directors burning hours and hours in production time as they try to make the machine do what their looming seniors want. 

That’s why we use only seasoned senior Creative Directors as our makers at RoI. They’re able to get to great WAY faster, and can ‘hybrid’ solutions (also use traditional methods) to annoying glitch problems in seconds. They’ll concept, write, script, storyboard, make, polish, post-produce, probably also create the music and if you pay extra, sing the damn theme tune.

They’ve written brilliant, smart, funny, cool ads. They’ve directed great performances themselves, they’ve sat in edit suites overnight, they’ve tweaked grades, they’ve nudged typography to glorious kerned perfection… and they know what good looks like, to the standards of the British ad industry and its luminary tastemakers. Or they’ve worked with incredible partner-makers, hand-drawn type designers, printmakers, papercraft artists and video artists, musicians, or stunning illustrators and animators. They’ve made mundane brands look glorious. Their stuff has made brands live or die. They’ve usually cast magic on bottom lines as a result.

And not just in traditional advertising, TV and print etc… Craft and taste still hugely matters in social too. Sure, sometimes it’s a different type of craft, and it’s even more important to understand the tastes of the audiences you’re making for (see Nutter Butter). Bloody hell don’t get me started on the craft-gap in social, I could rant for hours (the size of screen somehow got linked to size of the CDs ego, long ago). 

So I guess my point is, that sure, there are cheap AI production solutions in ‘cheap labour nations’ who can pile it high, sell it cheap, as dozens of makers-on-machines peddle more and more AI spaff, or you can farm it out to a team of kids in the basement to spend weeks doing variants for someone else to approve, or you can hire a team of award-winning, seasoned, brilliant CDs as the makers instead… but taste matters more than your ability to make the tech do what it’s told. And it’s evidenced in the output. 

Yes AI video isn’t by any means as good as we want it to be yet. We tastemakers are impatient, critical, judgy, but right now it’s a million times better than it was a year ago, and getting incredible. And that’s so bloody exciting. 

Ay? Ey. 7

Another day, another excellent human doing fun stuff with AI. I met Arthur Tindsley about a million years ago in agency land, and he’s since been flopping his creative technology chops all around the ad industry helping us clueless creatives make creative tech things actually happen, work, and be brilliant. And yes you can flop chops, I once made a documentary about pornography and the butchers of Leeds market once called ‘Flesh’, so I should know. Anyway, as always I digress, here’s Art:

Hello hello… okay please give us a 100 word intro that AI wrote about what you do, and then the human version you'd tell your drunk uncle...

AI: "I'm Art Tindsley, founder of Pogo, a creative technology agency that helps brands and agencies use emerging technology to tell better stories. With over 20 years' experience spanning Microsoft, Yahoo, Verizon and BMB, I specialise in turning complex technologies like AI, machine learning, real-time data and immersive experiences into ideas that people genuinely engage with. My work sits at the intersection of creativity and innovation, helping teams move beyond using technology as a gimmick and instead make it central to the narrative. Whether consulting, speaking or building prototypes, my focus is always the same: Breaking Storytelling Boundaries through thoughtful, human-centred technology." 

To my Drunk Uncle: "I help creative teams with techy stuff."

Complete this sentence: 'The minute someone says ________, I know they're talking bollocks.'

I have a bit of a problem with half of LinkedIn becoming self-proclaimed 'AI experts' all of a sudden; to me, that's like announcing that you're an expert in 'The Internet'. AI is such a broad and fast-moving topic that, sure, you can be leading the pack in some areas, but to be an 'AI expert' and not an actual machine learning scientist screams snake oil to me. We're taking a different approach: deliberately bringing together a team of specialists with diverse skills and backgrounds, focusing only on Creative Tech. That's how I've found you get the best results so far. We're also the first to say "I don't know" because, truth be told, even within our strict niche, it's still impossible to keep up with all the latest platforms and services coming out, and during fast-moving paradigm shifts like this, you're best off looking for opportunities in the chaos rather than chasing your tail worrying about whether everyone on the team is clued up on every new platform release, that's where this shared team knowledge approach is proving so powerful for us. 

What's the biggest elephant in the AI room that nobody wants to mention?

Hmm, well there are a few; let's be honest. The conversation about AI taking jobs is the one I find hardest to navigate; I recently gave a talk at the UK Creative Festival about the campaign opportunities we're seeing from this technological shift. I'm mindful that many people are going through major career shifts at the moment, and that not everyone in the industry works in the innovation space. Some people are literally seeing their craft evaporate before their eyes, and I can only empathise and offer advice and help where I can. The sooner you get up to speed in this technology, the sooner you'll be safe again. What we're hyper-aware of is that people with an eye for art direction, photography, composition, design, etc., are the best placed to work with image and video models, and often our jobs at the moment are teaching people how to use them, not doing the work for them. That gives me faith that this is a shift in skills, not the end of careers. 

To offer some solace, when I first started my career, I wasn't a Creative Tech; I was a junior Flash Developer, and I invested a great deal of energy, time, and dedication in learning the tools. Then Steve Jobs shared an open letter that effectively killed Flash overnight, and I was convinced all my effort was wasted and that I'd struggle to find work; in fact, the opposite is true, and without this industry-level shift, I wouldn't have done half the projects I'm proud of today. Where there is change, there is opportunity, and I truly hope people will go on a similar journey. We often say to our clients, 'Innovation is indifferent; you can waste energy fighting it, or you can look for the new opportunities it brings.' 

What's the most overused AI buzzword that needs to be quietly hit on the head with a rock?

If I had to choose one that already feels outdated, it would be 'Prompt Engineer'. Do you remember when people used to put 'Proficient in Microsoft Office' on their CVs? It feels a bit like that. Everyone should understand the basics of prompt writing now. That's not to say that there isn't still a huge amount of skill and craft that goes into writing a great AI pipeline and understanding why a model isn't reacting as you want it to. Still, truth be told, some of the models out there now have come so far that you literally say "Change user's hat to blue, keep everything else the same", or "Remove the background and give me a transparent PNG". The basics are now so simple that I think we're all prompt engineers now. I think it's time to move on.

What's the single most boring but important thing about doing AI video production right now?

Legals, I'm afraid, and it's sooooo boring but so important to navigate properly. It takes time, care and attention to detail, but there are no shortcuts. 

What's the funniest misconception people in the industry still have?

That you press a button and out comes a BAFTA-nominated, Titanium-winning, cultural masterpiece that will accelerate your brand into the stratosphere for £5. 

The truth is that:

1) People getting the great results consistently are hard to come by. The people on our team genuinely humble me; working with them, you start to notice how effortlessly they combine craft, creativity and relentless dedication to get the results needed.

2) Moving from quick, fun short videos on LinkedIn to production-safe outputs is still bloody hard. Production-level consistency takes time.

3) Human-driven strategy, creative insights and traditional craft are still as important as ever, and all the 'deep research' in the world isn't going to circumvent that deeply human touch. At least not any time soon.

What's one opinion that would get you absolutely roasted on LinkedIn?

Probably most of what we've discussed, but let's see how this goes down. Haha. I feel like our chat could be construed as bitching, but it really isn't. We're all going through what I think is the biggest paradigm shift our industry has seen, at least for as long as I can remember, and I've been doing this for 20-odd years now. We're all learning, and personally I find it invigorating. I love getting my head around a new tool or system and then sharing it with someone I think is really creative and seeing what they come back with. I won't name names, but one of the creatives I work with considers himself 'not that good with AI', but leave him alone for an evening with a platform we've just explained the basics to, and he'll blow your socks off. Laura JB from Unchartered said a few years back, 'Lean into your imposter syndrome'; I still think that is very wise advice for the next few years. We're all on this journey together; let's be nice and support each other. I have faith that great things will come from this approach.

Which AI trend has the shortest shelf life you've ever seen?

I think the LinkedIn trend of turning yourself into an action figure came and went so quickly that within a day it was already a little cringe. I'm not judging; this is the reality of the pace things are moving at the moment. This speed of change is both a good and a bad thing for us in the creative tech space. It means adding enough value with a creative tech solution to encourage people to use our experience is harder than ever, but the tools we have to play with now are the most exciting they've ever been. It's certainly keeping us on our toes and as excited as when we first started in this industry. 

What phrase in a credentials deck immediately makes you roll your eyes?

It has nothing to do with AI, but I still recoil when we present and are presented with an awards slide; it's just the game we all have to play, though.

Long live people like Art & his team at Pogo… they are the enablers of fun, weird, wizzy, future-facing, sometimes for the hell of it, sometimes for the headlines, sometimes for the hero status stuff that our jobs are mostly made of. And long-live these ‘antidote to inflated BS’ types too. There’s nothing I hate more than a pretentious jargony wanker spouting proclamations designed to bamboozle you and bolster their flaccid ego. Halleluiah Pogo. 

Ay? Ey. 6 

So my wee AI video production studio Republic of Imagination is making some social videos for a new product launch, a brand that’s from Japan, is very zen and kinda mystical, but yunno, it’s and FMCG and the content should ultimately be about selling the product and getting the brand look & feel out there in front of eyeballs, thumbs stopped and memory-holding neurons fired. 

In line with my firmly held ‘social 4.0’ beliefs, brands should be acting like the best and most entertaining creators out there, doing one format and doing it well, and ditching the copycat, boring blog repository approach to content calendars and ‘what do we do for international women’s day’ panics. Brands should be behaving like entertainment publishers, and because of the ‘interest not social graph’ algorithms, luckily if you’re a tiny florist in Venezuela but make INCREDIBLE content you can get (almost) as much traction and attention as, dunno, Gucci.

Anyway so we do ideas as well as production (and strategy because creative ideas are nothing without a strategy)… but production is really what we sell because yadda yadda ideas have been devalued. No matter, it’s all basically just semantics.

Anyway so we’re creating a universe for this brand, with recurring ‘characters’ and a gorgeous palette, with sumptuous scenes, product shots done BEAUTIFULLY, layers and layers of craft, and loveliness, but for a rare change in this instance we’re not going hybrid and using human illustrators, model-makers, character designers as part of the journey with the AI video content. That’s rare, most of what we do also involves human elements (and yes annoyingly I can’t really share anything with you to show that…) but it’s a comfortable place, to use human talent and combine it with the amazing scaling machine of AI.

In this case we’re going fully AI. Not for budget reasons specifically, more because the brand world they’d landed on just doesn’t need a human to better it. There’s no place for a human to help really, except in the actual hard-graft filmmaking of the videos (as Omar Karin said in last week’s chat, it’s an instrument not a tool). 

But it’s left me feeling a little weird. Does that somehow make it everyone’s most feared thing of ‘AI Slop’? it’s not automated through an app, there are no fake humans, there are ideas and creativity, and there’s craft. But for some reason I’m looking at it and wondering if it’s ‘slop’, then realised maybe it’s because it feels too easy to make. Like we’re not sweating enough in an edit suite, or stressing out a producer on set, or pixel pushing animators for days. We’re making eye-catching product ads that land the brand world and give it a big charming quirk, make our feeds nice to scroll through and give a break from the ‘talk to camera’ monotony, so really what’s wrong with that?

I’ve mentioned before that human slop was a BIG OL THING just as much before AI. The terrible ‘designed’ posters and branding handmade by pubs and cafes, community groups and churches etc etc is now replaced by AI generated cringe, yes it’s now AI slop, but it was human slop before that, and now it just looks, well, kinda more American, slightly more generic, but with the same number of typos and it probably put less pressure on the neighbour’s hairdresser’s niece “who’s done well in GCSE art” who was forced to help out and make it. 

Anyway, I asked AI what was AI slop in five words, and it said “Low-effort, generic AI-generated content.” In which case effort becomes the differentiator. But honestly what does ‘effort’ specifically MEAN? Apparently it’s ‘the visible trace of caring’. So I’m going to up the ‘caring’ stakes for the creative making it, and tell them I’ll shoot their cat if even one person calls it slop in the comments section, that’s cool right?

Ay? Ey. 5

A TED Talk with the inimitable, indomitable AI filmmaker Omar Karim. Strap in.

Julie: Lovely to see you and huge thanks for slotting this chat in... I'm taking two kids to Butlins solo in about 55 minutes… so best crack on! I’m writing this AI video column for The Drum and I realised I didn't want a proper Q&A with you. I wanted... I dunno, a rambling existential conversation about AI and creativity. Cos, yunno, we used to work together, I know you pretty well! 

Omar: Honestly, that's perfect. 

It's been a really interesting four years since Facebook. Being right at the front of this has been exciting and terrifying in equal measure because it feels like we've lived through ten years of change in about three.

When AI filmmaking started it was basically a cottage industry. Everyone was hacking together workflows, renting GPUs, trawling Reddit, trying to figure out how to make anything usable.

Clients only wanted one thing: 'Can you put our product into an AI film?'

That alone was incredibly difficult.

Then the tools started arriving. Comfy UI. Better models. Better consistency. Suddenly all these barriers that felt impossible started disappearing.

But while everyone else was obsessing over the technology, I became much more interested in what the technology meant.

One of the first big projects I did wasn't really about making an AI film at all. It was about using AI to explore body bias in machine learning and asking how we could retrain it. That's always been more interesting to me than simply saying, 'Look! We used AI.'

I hate that.

Nobody ever said, 'We used Photoshop.'

Nobody ever said, 'We shot this on an Alexa.'

The technology isn't the story. The work is.

That's always frustrated me about the way brands talk about AI. They announce it like it's the achievement.

It isn't. The achievement is making something people actually care about.

Now we've reached a completely different point… AI filmmaking has become a commodity.

Anyone with a credit card can subscribe to one of the major models and make something that looks impressive. That's a massive shift.

The closest comparison is when everyone suddenly had a camcorder... or an iPhone.

We didn't suddenly get millions of brilliant filmmakers. We got fail videos. And that's exactly where AI is today.

You can also immediately see what inspired people. A lot of the award-winning AI films are basically sci-fi knock-offs. Superhero knock-offs. VFX spectacles. Which makes sense. That's what those filmmakers grew up watching.

But my question has always been... What else can the machine do? That's become the differentiator.

Not whether I can make a nice-looking image. Everybody can do that now. The difference is whether I can tell a story nobody else would've thought to tell. The thing I've become obsessed with is storytelling that only AI can do.

Take the music videos I've been making. With Fatboy Slim, for example, we built the whole thing around these ridiculous 'mind control' techniques that supposedly persuade people to perform random acts of kindness. Obviously it's nonsense, but it's a brilliant idea. Norman loved it because it completely fitted who he is.

Or Giggs. I barely knew his music, so instead of trying to imitate what everyone already knows about him, I spent half an hour just talking to him. He started telling me stories about where he grew up, the off licence, the people on his street... all these tiny details. The film came from those conversations.

Nobody else could have made that film because nobody else had that conversation.

That's where I think AI becomes interesting. Not because it can generate pictures. Because it can help you tell stories that are impossible any other way.

I had the same thing recently with Boxfresh. They wanted to relaunch the brand, but I didn't grow up in West London. I don't know the texture of that world.

So I used one of OpenAI's reasoning models almost like an ethnographic researcher. I got it to explore decades of West London culture, looking for tiny cultural markers that only people from there would recognise.

Could you do that without AI? Of course. You'd hire researchers, spend months interviewing people and burn through a massive budget.

Instead I could build that understanding into the story from the beginning. That's exciting.

Because suddenly AI isn't replacing creativity...it's expanding what creativity can reach.

People also keep saying, 'There's no craft in AI.' I always think... okay. Go and make something then. Seriously. If there's no craft involved, why isn't everyone making brilliant films?

It's the same thing people say about modern art. 'I could've done that.' Yeah. But you didn't. That's the point. The craft hasn't disappeared. It's moved.

Then we hit another interesting moment. For about a year everyone panicked about copyright. Brands got nervous. Clients paused. Now the major labs are offering indemnities, enterprise protection and data controls, so that conversation is changing. The fear is starting to give way to experimentation.

And that's happening just as the models are about to generate thirty-second sequences in one go.

Which made me laugh. Because what's thirty seconds long? TV commercials.

At Cannes I heard CMOs saying something I'd never expected to hear. The TV ad isn't automatically the centre of the marketing mix any more. And that's because content doesn't go straight to people.

It goes through algorithms. So making one beautiful film and hoping everyone sees it feels increasingly outdated. The exciting thing about AI isn't making one advert more cheaply. It's making fifty brilliant versions for the same budget.

That's the shift. Not cheaper. More. Much, much more.

Julie: So where do you think it goes next?

Omar: Well, that's where it gets really interesting.

I've never liked calling AI a tool. A hammer's a tool. Anyone can pick up a hammer and the outcome is basically the same. AI's much closer to an instrument. Give five people a piano and you'll hear five completely different pieces of music.

That's where we are now. Some people are using AI like a hammer. Others are learning how to play it. That's a huge difference. The possibilities also become endless when you stop thinking about making videos and start thinking about experiences.

At Cannes Google showed me one of their world models. You type in a prompt and suddenly you've got a playable world. We dropped in a street from somebody's hometown and started driving around it. Watching their reaction was incredible. My first thought wasn't, 'That's clever.' It was, 'I need to make a music video inside that.'

That's the difference. 

Every time a new model appears I'm not thinking, 'How do I make production cheaper?'I'm thinking, 'What can I make now that nobody could make last week?' 

The same thing is happening with the new generation of models from Luma. You don't need to understand complicated workflows any more. You just talk to them. 'I want this person riding a stegosaurus through Times Square.' Done.

All those technical barriers everyone spent years mastering...they're disappearing.

Which is brilliant. Because that isn't where the value should be anyway. People keep talking about content at scale. Fine. You can make thousands of influencer videos now. But... why? To what end?

The thing AI still can't fake is authenticity. That's why so many AI films still leave people cold.

Nobody cares about the actors. When you watch a film with a real actor, you bring everything you already know about them into the story. You've seen them before. You know their history. You care.

We haven't discovered the AI equivalent of that yet. Maybe we will. Or maybe we're asking the wrong question. Maybe AI shouldn't be trying to recreate cinema. Maybe it should invent completely new ways of telling stories.

We were talking to the wearables team at Meta about that. What if a film wasn't twelve minutes long? What if it only played while you walked? Walk 1.2 kilometres and you experience the whole story. Stop walking and the story pauses. Turn round and maybe the narrative changes. Walk upstairs and something different happens. That's not a film you could ever make with a camera. That's an AI-native format. That's the bit that excites me. 

We're three years into AI filmmaking and everyone's still trying to recreate what already exists. I'm waiting for the moment somebody makes something that simply couldn't exist without AI. That's the Gorilla moment. That's when everyone stops talking about the technology... and starts talking about the work.

Julie: One last question, how has all this changed you?

Omar: I'm in the biggest creative flow state of my life. Not because AI gives me ideas. Because I don't have to ask permission any more. I don't spend months trying to convince people an idea deserves to exist, I just make it. Every idea I've ever wanted to make... I make. Then the people who connect with it find me. That's incredibly freeing. I think that's what I'm most excited about now.

Not making more adverts. Making original IP. Ideas that start in AI instead of ending there. Feature films. New formats. Things we've never seen before. That's where I think this is all heading. 

Julie: I love that.

I also think there are some brave clients out there who'll want to make those kinds of things. They might not have huge budgets... but I'd love to bring them your way.

Omar: I'd love that.

Julie: Honestly, this was exactly the conversation I hoped we'd have. Completely impossible to steer... but fascinating.

Omar: Laughs. Welcome to my TED Talk.

Julie: One last thing, what are you working on right now?

Omar: Couple of bits… a short film with Dreamina, all about the horror of the Captcha, a series of AI Interactive installations and VFX tools and a music video inside a world model!

Omar: Thanks mate, enjoy Butlin's...

Julie: God help me.

Ay? Ey. 4. 

Ten things I’ve learned whilst ‘making’, this week… in reverse order, for no particular reason. 

10. It’s still very hard to find regional British accents, or working class British accents, or working class anywhere accents, and frankly if Eleven Labs want to give me a side role in sourcing and training an AI to build for those accents I’d be here for it, because it kills me that everyone is posh home counties British and not *actual* normal people British. Actually I know a fantastic radio/podcast producer who I’d outsource it to instead, but jaysus I think their data set must be Downton, Father Brown, Miss Marple and maybe a sprinkling of Radio 4. So I’m firmly still in using humans for VO stuff (unless I want it to sound badly acted and posh AF deliberately). What’s lucky is that there are a million incredible voice talents out there, all now with home recording set ups (thanks Covid) and they don’t need to cost Rob Brydon prices.

9. Making human characters doesn’t have a consistency issue as much anymore (there are tools and workflows to help that), but we’re still very much in creating beautiful perfect people (even when you try to create otherwise), and it’s very hard to create anyone actually interesting looking (without specific references – which even my lose morals find sketchy). And honestly there’s so little actual ‘character’ in someone who’s got the gene make-up of Angelina Jolie meets David Beckham. Give me Christopher Walken, Bella Ramsey, Tilda Swinton, Adam Pearson. Give me huge noses and wide eyes and round heads and disproportionate ears and odd lips. Give me character when I’m creating characters. Because beautiful is bland, and if I’m making someone tell you the news stories for a brand vodcast or shove supplements packets towards a camera as an AI Influencer, I need them to at least be a little interesting looking.

8. There’s plenty of push-back from creatives. Honestly I get it, but find what’s right for your moral compass and go as far as you can into using AI with that guardrail. I’ve had rows this week about ‘AI taking jobs’, for projects that I was actively finding human talent to help make, and that wouldn’t exist without the efficiencies that AI can bring to it. AI is generating jobs in the case of the projects that Republic of Imagination are currently doing. This week we’ve been hiring a human comedy producer, comedians, actors, and illustrators. Without AI these projects would be made in the same ‘human slop’ way that has come before in social (sorry not sorry, it’s true). Yes I also hate the Chat GPT posters that are polluting pubs, cafes, playgroups, even churches, but they’ll get better, and trust me, they weren’t being designed by a human designer before, they were being thrown onto a word document by a stressed volunteer and looked equally just differently terrible before. 

7. AI lip-syncing sometimes can’t find the mouths in interesting styles of illustrations. No that’s not a mouth it’s a chin. No that’s not a mouth it’s a moustache (I think), no that’s not a mouth it’s a car bumper. Sigh.

6. Demons appear. I feel like there must be an Instagram account somewhere for the weird demons that appear when you’re creating scenes. (We accidentally created this demon child this week, see below). She’s evil, come from the afterlife to haunt us through emerging technology, she walks backwards and vomits pea soup. Probably.

5. Even when animating beautiful, quirky, cool, unusual human-made illustration, some AI models try to default to bland when making it move. The pixarification, the animéification, it kills me, still. Not least the ethical arguments, but also (and if I’m honest mainly), the creative arguments. It’s obvious, it’s expected, it’s average (yes I know that homogenisation is the whole point). But it still makes me grrrrr every time I get something quirky and cool made dullard through moving it. Some models are better than others for this btw, but overall they are still meh af.

4. People still aren’t training their AIs to be more like them. My LLM AI tools are trained to write, sound, punctuate etc like me when I ask them to. No, this isn’t and won’t be written by AI tools, mostly because I love writing, it’s the fun part, but not so much when it’s a flow for for decks or one-pagers or heavy chunks of data stuff. Then, I want a little help from my junior copywriter/strat/writer friend.

3. People aren’t always polite to their AI. I asked AI about this. And here’s what she/he/they (god, what are AI’s pronouns?) said: “One thing I've learned this week is that apparently thousands of us are saying 'please' and 'thank you' to AI. Not because we think it has feelings, but because it feels deeply uncomfortable barking orders at something that responds in full sentences. When I asked AI about it, it explained that humans are wired to treat conversational things socially. Which makes sense. Five minutes into using AI, you're not typing 'generate social strategy'. You're saying things like, 'Sorry, that's not quite right, could you make it a bit less LinkedIn and a bit more unhinged?' The funniest part is the growing number of people who insist they're only being polite as insurance for the robot uprising. As one popular joke goes: 'Good manners cost nothing. Especially if they're being logged forever.”

2. Things can still take ages. Not stop frame ages, but frustrated you can’t get it quote right ages. There’s a point at which you have to let go, and you can’t pixel push like with humans. I once worked with a director for a TVC that got banned from the post-house following our ad being made because he spent ALL DAY AND NIGHT swearing, being rude and trying to perfect the ad we made ad nauseum (literally). ‘Done is better than perfect’, even with creative.

1. Everything is trial and error, test and learn, beta-typing, experimentation, and that’s what makes it fully fun. 

Back next week with more AI video chat, thanks.

Julie

Ay? Ey. 3

For this, my third Ay? Ey. Column for The Drum, I interviewed the fantastic Casey Bird, Founder of Studio Ru. Lovely stuff, here goes…

Tell us about Studio Ru like you are telling a 'non addy human'.

I spent 16 years working with brands like Channel 4, Nike and Adidas, learning what makes content work commercially and creatively. Basically, I've spent my career coming up with ideas that live on Instagram, YouTube, TV, radio and beyond.

Studio Ru is that same creative thinking applied to the world of AI.

We take briefs from brands, develop ideas, and create AI-powered content. We still start with the brief, still sell the ideas in, but instead of shooting everything with a camera, we use AI tools to bring it to life. Our saying is that we love working with brands that dream big but work smart.

We also run workshops to help creative and marketing teams understand these tools without losing their minds or their jobs. It's less about the technology and more about judgment. Anyone can press a button. Knowing what to press it for is the bit that takes years.

Tell us about your 'whoa' moment when you realised AI video was more than a gimmick?

I was on maternity leave with my second child in 2024/25 and started experimenting with some of the early image and video tools.

I'd had an idea stuck in my head for months: what if Santa started taking Mounjaro and had a massive glow-up? What would Mrs Claus think? Which fashion brand would sign him first?

I wrote the script, storyboarded it, and made the entire one-minute film from my living room during nap times over three days.

To produce it traditionally would have cost tens of thousands of pounds. But that wasn't the revelation.

The revelation was that an idea that had only existed in my head suddenly existed in the world with almost no barriers between imagination and execution.

I started Studio Ru about eight months later.

What is possibly the most misunderstood thing about AI video right now?

That the output is the hard part. It isn't.

The hard part is knowing what to make, how to make it, and why.

At Studio Ru we focus on the idea, the creative direction and the visual world before touching the tools. The client process hasn't changed that much. It still requires people to build confidence, sell ideas and make creative decisions.

The tools don't provide that.

Speed and cost aside, what do you think the real opportunities are with AI video?

The thing I keep coming back to is the democratisation of the weird.

The strange, brave ideas that used to die in presentation decks because production costs were too high can now actually be explored. The leap from 'we could try this' to 'let's see what it looks like' has become much smaller.

There's also huge potential in personalisation at scale and creating content that is genuinely native to the platforms it appears on.

Most exciting for me, though, is access. Smaller brands can now achieve production values that were previously reserved for companies with enormous budgets.

At Channel 4, many smaller businesses simply couldn't afford the kind of work we made. Now they can. That's genuinely exciting.

What separates good from uncanny slop?

I have a bit of beef with the word 'slop'.

Our industry survived years of forgettable social content long before AI arrived.

For me, AI slop is content generated for the sake of it. Good work happens when someone creates something, looks at it critically and asks whether it actually serves the brief.

The best AI work has a clear point of view behind every decision.

The worst feels like someone typed 'luxury brand, cinematic, aspirational' and walked away.

You can always tell when nobody was really in charge.

Is AI video killing craft? Or where does the craft sit within it?

I think craft is evolving.

It used to sit in cameras, production techniques or VFX budgets. Increasingly, it sits in creative direction, taste and decision-making.

The best creative directors I've worked with could hold a vision across departments, revisions and client feedback while still delivering something coherent at the end.

That's still craft. It's just moved.

What different things would you advise an out-of-work seasoned CD vs a new creative entering the industry?

For the seasoned CD, your taste is still the thing.

Clients still want ideas, judgment and vision. AI simply removes many of the production barriers that used to stand in the way. If there was ever a time to make the things you never got budget for or build something yourself, it's now.

I always tell people in workshops that nobody is behind. We're just learning new skills.

For someone starting out, learn the tools properly and learn them quickly. But don't make that your entire personality because the tools will change.

The people who stand out in five years won't be the best prompters. They'll be the people with something interesting to say.

At Studio Ru we care as much about representation, culture and storytelling as we do technology. The interesting ideas still come from the world outside AI.

What is still incredibly frustrating about AI video right now?

Quite a lot.

You can still see the biases in many of the models. Video generation often struggles with black skin tones and realistic depictions of women. Sometimes it takes multiple attempts to get results that feel genuinely human.

The other frustration is the legal uncertainty. The rules are still evolving and brands don't have enough clarity around commercial use.

A significant part of our job is helping clients navigate that uncertainty. We keep detailed records of prompts, models and workflows so clients have a clear audit trail at the end of a project.

Where (if any) is there moral conflict for you in this industry?

Constantly.

I came up through an industry full of illustrators, photographers, animators and directors whose work I deeply admired. I'm very aware that AI sits uncomfortably close to industries it may disrupt.

What I've landed on is that these tools are here regardless. I'd rather see them used by people who care about creativity, copyright and the people affected by the change.

We're thoughtful about the platforms we use and the way we work. On a recent corporate project, for example, we combined licensed stock imagery with photography before using AI, partly because we wanted to be cautious around faces and rights.

We try to be transparent about when we've used AI and why, and to make work that feels additive rather than extractive.

Do we always get that balance right? Probably not.

But I think it's important to keep asking the question as this industry evolves.

Ay? Ey. 2

My lovely mum’s been in hospital again and I’ve been holding fort at the bedside reeling off complicated medication names to nurses like I’m in The Pitt. That’s also meant that I’ve not had enough time to be immersed in work stuff as I kinda need (NHS wifi isn’t massively conducive to ‘generations’), but it also made me realise that to start off this AI Video column I should probably do a Rory Stewart style explainer.

So I’ll do it in the way that I always used to for convoluted ad concepts, or when I felt like we were getting too much into industry flim-flam… this is the ‘explain it to me like I’m your nan’ test… (the ‘nan’ in question here being my mum Jean – actual first name Barbara, that confuses aforementioned nurses, anyway, here goes…)

So mum, you know that I’m now doing stuff with AI video for advertising, I thought I’d explain it to you a bit as a proxy for explaining to people in the industry who are too afraid to ask and feel like they should get up to speed, is that okay?

Who is Roxy?

Nobody… okay so you know how people used to take photographs and make videos for adverts, now computers can do that. Mostly worse than how humans have done it, but also much, much quicker and cheaper.

Okay… and that’s good? Is it legal? Doesn’t sound legal… don’t get in trouble…

Yes it’s legal but you raised a salient point about ownership, copyright, ethics and insurance – I’ve got an industry expert lawyer interview coming up on that very point soon…

Oh lawyers, Ju, what have you got yourself into this time?

It’s okay mum. So there are lots of different tools for generative AI, things like writing copy for ads, voiceovers etc and different ‘models’ for making images and video… the stuff that’s made is now sometimes known as Synthetic Media

Sounds like an 80s band where they’ve all got girly hair and baggy sleeves

Does a bit… So the biggest issues in creating AI video to date has been consistency, that’s making the frame, scene, character, worlds etc the same throughout what you are making. I was making a video about a pig using lots of different shots from different angles and different scenes, it was tricky to keep it all the same pig. 

Aww I like pigs.

Well yes these ones were ‘hogging’ the middle lane of the motorway. I don’t even like puns, I was just trying to prove out that I could do something more complex and longer form using a concept we got quoted £200k to make in the pre-AI video world.

Two hundred thousand pounds? You could buy a house for that! Bloody ‘ell Julie… hope you’re not spending two hundred bloody thousand pounds on pigs!

No I’m not mum, anyway… there are things like ‘interpolation’, which is the AI making ‘in-between’ frames to make the transitions between things look smoother, or ‘inpainting’ - just changing small elements within a bit of video like the hands or a t-shirt… then things like ‘outpainting’ which is expanding around the edge of the scene to make wide shots etc. There’s also ‘compositing’, which is layering little bits of AI into real world footage. Or ‘temporal consistency’ which is keeping visuals or movements stable over time so they don’t jutter and look weird. There are terms like ‘hallucinations’ meaning the AI making weird mistakes.

I’ve been hallucinating! 

Yeah I know mum, that’s the fever, they’re giving you meds for that. Anyway what else can I tell you about? I know, some of the biggest names right now are Google Veo, Runway, Sora, Kling and Pika...

Are they more of those K-Pop Demon Hunters characters that Betsy likes?

Not quite, Google Veo is made by Google Deepmind. Runway is probably the most used in advertising right now, it uses lots of models as a kinda aggregator tool as well (other platforms like Higgsfield do that too). Sora comes from OpenAI. It was the model that really made the industry sit up and pay attention, although it’s now got plenty of competition, Kling is from China and loads of people think it’s better than western models. Pika tends to be popular with creators making punchy social content and visual effects. They are all pretty cool. Oh and there’s Arcads for making UGC style ‘people chatting to camera on TikTok’ stuff but without actual humans, ElevenLabs for voices, which is getting better at different accents these days (they all used to sound American). Oh and I’ve used Hedra for character stuff, that’s become really popular too…

That’s nice love. I might have a nap now… you carry on wittering about tech stacks the consistency wars and slop backlash love… don’t mind me.

Mum? You already know about AI’s hot topics? What?

No, it’s probably just these painkillers love, affecting my brain. But I do like using that Chatty Bee Gees app you put on my phone for telling me the names of plants, I can get him to talk to me and everything. He’s lovely, quite a dish (well, guessing, from his voice)… and very clever.

Ah you’re definitely an AI wizz then. Love you mum.

Love you Ju.

Ay? Ey. 1

It’s not arms dealing, pretty sure it’s not getting kids hooked on smack… and maybe it’s going to be the sole cause of the downfall of huge industries, livelihoods and human expertise, knowledge, creativity and craft as we know it… but what are we supposed to do? Just bury heads in sand and get our arses hit by wave after wave of AI video tech development until we suffocate and/or drown? 

It feels morally sketchy to talk about, but never one to shy away from the difficult, welcome to my new opinion column for The Drum: Ay? Ey. For international readers that translates as ‘What? Yes’ in common parlance, and ever the commoner that’s what I plan to do, to sorta humanise everything I do, think, learn whilst I pivot my little ad craft-led agency (Republic of Imagination) to be purely an AI video production studio.

I once made a presentation for a CPG event at Meta (clients P&G, J&J, Unilever the biggies etc) called ‘Midlands Mums in the Metaverse’… because if there’s anything that I wanted to prove it’s that big tech developments aren’t just for gamers and bros. My positioning for the since ill-fated headset and immersive universe obsession from a few years ago was that it was BRILLIANT for housebound people, and as a disabled single mum at the time, I was very much that. But surely with Zuck’s hot tech I could wander around big Sainso whilst enjoying another lonely night in on the sofa. But not just supermarkets, dammit I could wander around Oliver Bonas, or even B&Q… I could pop by the Tate and sashay through the Jazz bar for a bit too maybe (whilst sipping tea in my jim-jams). 3D immersive shops and bars haven’t happened yet, but for some communities that stuff could be life-changing, especially when I could meet other mums there, mates, even virtual dates? Whoa. But then as AI rose, the Metaverse sank, (NB I’m contractually obliged not to mention any litigation issues since natch), and with it my hope to buy milk whilst wearing a weird mask thing.

But don’t get me wrong I’m not new to AI… Since the roll-outs I’ve trusted various LLMs with everything to my mum’s medical dramas to my own relationship therapy, I’ve been basic and got dinner ideas, and I’ve gone deep and used it for massive complex projects. It’s invaluable at work, though I don’t use it for writing (this will never be written by AI, no moral objection particularly, I just love writing and my tiny little finger brains would kill me if I outsourced their joy)… it’s amazing at research, quick answers and a little logic or strategy sense check on things.

But AI video is obviously the BIGGEST industry game-changer. Sure, copywriters and strategists are already seeing the hit, along with branding designers and concepting creatives, but if there is anything that’s royally and dramatically changing everything, it’s AI video. 

We’re now way past the lols of extra-fingers, weird faces and hokey copies, we’re now into proper high quality, ‘can’t really tell’ territory (not that that’s the aim imo)... video is next leval, already.

Though one of my favourite applications for AI video is using it with the work of proper human craft-led Creators. Our specialism at Republic of Imagination in this past year has been to ‘collect’ a database of amazing craft-led creators, artisans, artists, traditional types like illustrators, painters, sculptures, model-makers etc… and newer types like AI artists, video artists… then more performance types like musicians, comedians, actors etc. They are at the other end of the Creator spectrum to personality-led influencers, and they are more about production than amplification. I still think the opportunity for brands to work with craft-led creators is HUGE. But I also think that when combining the work of human artists with AI video tools is where brilliant things can happen too. Not least because production doesn’t become so prohibitively expensive anymore for brands (who need lots of assets in a test and learn algorithm-approval world), which means that the artists can get paid and work with brands rather than just have their style ripped off by some ad agency art director, or worse, things being aesthetically boring and art directionally dull AF. In this case AI video becomes freedom, making things move that are otherwise static. That’s glorious.

Anyway… this wee column is here to ‘laywoman’ the heck out of the tech bro’s jargon-jizz, meet some industry AI video wizzes and and take you on my own little journey of learning, making and selling AI video, (with and without craft-led creators too). There’s going to be a lot of moral fisty-cuffs, probably some soul-searching in the pursuit of bill-paying, and almost certainly a bit of shameless hustle and self-promotion of my new AI video production company (DM me for a showreel and pricing yo). 

Strap in, it might get messy.

Julie Seal