We use AI. We still film everything that matters

How a Bristol production studio is folding generative tools into real production and where a camera still wins.

There’s a tidy version of this conversation where a production company plants a flag: AI is the enemy, craft is sacred, real cameras forever. It’s a comforting story, especially if you own a lot of cameras. It’s also not quite honest.

So here’s the honest version. We use AI in our work now. For images, for motion tests, for the unglamorous middle of production that clients never actually see. And we still put a crew, a lens and real people in a real place for anything that has to make someone feel something. Both of those things are true at the same time, and pretending otherwise doesn’t help anyone with a campaign to brief this year.

Where AI genuinely earns its place

The most useful thing about generative tools right now isn’t the finished film. It’s everything around it.

In the pitch, we can show you a concept instead of describing it. Moodboards, style frames and rough motion that used to take days now take an afternoon, which means you’re reacting to something real before a penny of budget is committed. That’s not a gimmick, it’s the difference between approving a direction on faith and approving it because you’ve seen it.

Then there’s the work nobody romanticises: extending a background so a shot works in both 16:9 and 9:16, cleaning a plate, reducing background noise, generating a placeholder while the real asset is still in the diary, spinning up twenty variations of a paid-social frame to test which hook actually lands. This is where AI quietly saves hours and money, and where clinging to “we shoot everything” would just be charging you for our principles.

Where a camera still wins…and it isn’t close

Here’s the part the AI-hype cycle skips. The moment a piece of content has to earn trust, real production pulls ahead and stays ahead.

Real faces your audience believes. Real light on a real product. A founder who’s visibly nervous and therefore visibly human. The texture of a place that actually exists. AI is remarkable at plausible; it is still poor at true. And audiences, especially the ones who’ve spent two years watching the internet fill up with uncanny, slightly-too-smooth imagery, can feel the difference even when they can’t name it. For brand films, for anything with a person’s reputation attached, for the B Corp clients we work with whose whole promise is honesty, “plausible” isn’t good enough.

The bit the industry doesn’t like saying out loud

None of this arrives without cost, and we’d rather be straight about it.

Some roles are changing under our feet. The junior retoucher, the stock-footage researcher – those jobs are genuinely shifting, and it’s disingenuous to wave that away with “AI just frees you up to be more creative.” It sometimes does. It also sometimes just erases the entry-level jobs where people used to learn the craft in the first place. The studios that come out of this well will be the ones that retrain people into direction, judgement and taste – the things a model can’t hold – rather than quietly cutting the headcount and pocketing the difference.

There’s an environmental line to walk too, and it’s worth putting real numbers on it rather than hand-waving. Generative compute isn’t free. Making a single five-second AI video on a current top-end model burns roughly 3.4 million joules – about an hour in a microwave, or 38 miles on an e-bike (MIT Technology Review, 2025). A still image is far smaller, roughly the energy of charging your phone, and a text prompt smaller still: a thousand of them cost about 16% of a phone charge (Hugging Face, 2023).

Now weigh that against a shoot. Flying one crew member to an overseas location and back runs to somewhere around a tonne of CO₂ per person – a meaningful slice of an individual’s entire yearly footprint. In pure per-asset terms, a single long-haul seat dwarfs thousands of AI generations, and it’s dishonest for anyone in our industry to pretend a location shoot is automatically the greener choice.

But we won’t hide behind that comparison either, because it flatters AI. A finished project is never one clip – it’s hundreds of iterations, re-renders and dead ends, and the data centres behind them are among the fastest-growing energy loads on the planet. A shoot’s footprint is visible and front-loaded; AI’s is diffuse, cumulative and easy to pretend isn’t there.

As a B Corp, our job isn’t to crown a winner – it’s to actually run the sums for the brief in front of us, and sometimes the lower-impact answer really is a laptop rather than a long-haul flight.

What this actually means for your next brief

Stop asking “AI or a shoot?” It’s the wrong question, and it usually comes loaded with either fear or hype. The better question is: what does this specific piece of content need to do?

If it needs to move fast, test cheaply, or handle the invisible plumbing of a campaign, AI probably belongs in the mix. If it needs to make someone trust you, feel something, or believe a human stood behind it – that’s a shoot, and we’ll happily talk you out of the shortcut.

That’s the whole approach, really. We didn’t pick a side. We picked the tool that serves the work, kept the craft where craft matters, and stayed honest with clients about which is which. Turns out that’s a more useful place to stand than either end of the argument.

Skylark Media is a B Corp-certified production and creative studio in Bristol, making everything from TV ads to social content. If you’re weighing up how AI fits your next campaign, we’re happy to talk it through.