By Angus Mead, Founder, Coral Film Co – video production on the Sunshine Coast and across South East Queensland
There are two conversations happening about AI in video right now, and neither of them is honest.
The first says AI is about to wipe out video production. Type a prompt, get a commercial, everybody who owns a camera is out of work by Christmas. The second says it’s all slop, it will never look right, and real filmmakers should ignore it.
We sit somewhere in the middle, because we actually use this stuff every week on paid client work. Coral Film Co. has been producing video on the Sunshine Coast for over ten years, for builders, developers, hospitality operators and national brands. We run Higgsfield as our main AI platform, which gives us access to models like Kling, Seedance, Gemini and Google Veo. We’ve burned a lot of credits, produced a lot of garbage, and produced some genuinely brilliant work we could not have made two years ago at any price.
So here’s the field report from a working production company. What’s useful, what’s a money pit, and what we think takes over next.
What AI is actually good at in 2026
Bringing still photos to life
This is the big one. If we had to keep one AI capability and throw the rest away, it would be image to video.
Feed it a well composed still, give it a considered prompt, and you get a living shot. A slow push in. A gentle parallax. Light shifting across a surface. It’s the closest thing to free camera movement that has ever existed.
For architectural and property work it’s a genuine unlock, and on the Sunshine Coast that’s a huge slice of the market. We use it for transitions for builders, moving between a render and a finished home, or drifting through a space we only ever photographed. Builders and developers who never had the budget for a jib, a slider and a two day shoot now get motion in their marketing. That’s not a gimmick, that’s a new product line.
Relighting, and never fearing the weather again
This one changed how we quote.
Every video production company on the Sunshine Coast knows the feeling. You’ve got one shoot day booked in Noosa or Maroochydore, the client’s schedule is locked, and you wake up to flat grey cloud rolling in off the water. Historically that was a disaster. You either shot it and delivered something lifeless, or you paid to come back.
Now we can take that flat grey footage and turn it into a golden afternoon. Warm rim light on the fascia, long shadows across the driveway, the kind of moody coastal sun you’d normally wait a week for. It isn’t a filter. The models understand where light should fall and how a surface should react to it.
The commercial implication is bigger than the creative one. Weather risk used to be a real line item and a real cause of stress on every South East Queensland shoot. It’s now largely a post decision.
Photo editing and lighting elements
Adding a practical light. Cleaning a frame. Fixing something that got missed on the day. Building the plate you actually wanted rather than the one you got. The generative photo tools are mature, fast, and cheap compared with the video side, and if you’re pairing them with image to video you’re effectively art directing a shot after the fact.
Rotoscoping
Deeply unglamorous and one of the most valuable things on this list. AI roto has taken tasks that used to be a full afternoon of frame by frame misery and turned them into something you set running while you make a coffee. It’s not always perfect and it still needs a human to clean up the edges, but it has quietly removed the worst hours of the week from our post schedule.
Anything that gives an editor back an afternoon pays for itself immediately.
Where it still falls flat on its face
It cannot edit your video files
This is the single biggest misunderstanding people have. AI is superb at generating video. It is close to useless at editing the video you already shot.
Give it a 90 second interview and ask it to cut, pace, structure and shape it, and you’ll get something that technically follows instructions and is emotionally dead. It doesn’t know that the good take is the second one where the guy laughs. It doesn’t know to hold on the silence. It doesn’t know that the story lives in the bit that looks like a mistake.
Editing is not an assembly problem. It’s a taste problem. That gap is still enormous.
Renders can’t be fixed, only re-rolled
Here’s the thing nobody tells you until you’re paying for it. A render is a black box. If the hand is wrong, if the car warps, if the light does something strange in the last 12 frames, you cannot go in and fix it. There are no layers. There is no timeline. There’s no keyframe to nudge.
Your only option is to change the prompt and roll again. And again. And again. Which brings us to the money.
It is expensive, and the cost is hidden in the failures
Credits vanish faster than you can believe. The cost per usable second is not the cost per generated second, and the gap between those two numbers is where new users get destroyed. Ten generations to get one shot you’d actually put in front of a client is normal. On a hard shot it’s worse.
Anyone quoting AI video work as if it’s cheap has either not done much of it or is about to lose money on the job. Budget for iteration or don’t budget at all.
The prompt is the craft
You will hear that prompting is a fad. It isn’t, at least not yet. There’s an enormous difference between what a first time user gets out of these models and what someone gets who understands how to phrase camera language, how to describe light, how much to specify and, critically, when to shut up and let the model work.
Bad prompt, bad output, every single time. The models are not mind readers and they reward specificity in some areas and punish it in others. That knowledge only comes from burning credits and paying attention.
The artistic decisions are all still yours
This is the part people skip past. AI has opened a lot of creative doors. It has not walked through any of them for you. Every decision that makes a video good, what the story is, what the tone is, what stays and what goes, where the cut lands, why the client’s message is actually the wrong message, is still a human sitting there with an opinion.
The click of a button is a long way off. Not months. Years.
How to actually capitalise on it
If you’re running a production company, or you’re a business weighing up who to hire, here’s how we’d approach it.
Sell outcomes, not AI. No client cares which model made the shot. They care that their house looks incredible in the rain. Don’t lead with the tech, lead with the result. The moment “AI” becomes the pitch, you’ve invited a price conversation you don’t want.
Build products around the things it does brilliantly. Motion from stills is a productised service. Photo library to hero video for builders, real estate agents, hospitality venues, anyone with an archive of images and no video. That’s a real offer you can put a price on tomorrow.
Price for iteration. Bake failed renders into the quote. If you quote as though every generation lands, you’re funding the client’s experiments out of your own pocket.
Use it to de-risk shoot days. Fewer contingency days, less weather anxiety, fewer reshoots. On a coastal shoot schedule that saving is real, and you don’t necessarily have to hand all of it back.
Use it as a pitching weapon. Previs, mood pieces and concept films that used to be a written treatment can now be an actual moving thing. Turning up to a pitch with a version of the idea, rather than a description of the idea, wins work.
Do not fire your editor. Seriously. The value of taste goes up, not down, as generation gets cheap.
What we think takes over next
Editable generations. The first platform that gives us layers, masks, regional re-rolls and the ability to fix a single element without re-rolling the whole shot will take the entire market. That is the missing piece, and everyone building in this space knows it.
Persistence. Character, product and location consistency across shots is the wall between “cool clip” and “actual film”. It’s improving quickly. When you can hold a face and a room across 20 shots reliably, narrative work changes overnight.
AI as a layer inside the edit suite, not a separate destination. Roto, relight, cleanup, extension, upscaling and dubbing living natively in the timeline. The standalone generation platform is a transitional product. The endgame is the tools we already use just quietly getting superpowers.
A collapse in the price of generic content, and a premium on the specific. Anything that could be described as “a video like the other ones” is heading toward free. Anything that depends on a real person, a real place, a real story or a real point of view holds its value, and probably gains it.
The bottom line
AI in 2026 is a genuinely great tool and a genuinely terrible employee.
It has opened creative doors we were locked out of a couple of years ago, it has removed some of the worst grunt work from post, and it has made the weather a lot less frightening. It has also cost us a fortune in failed renders, it cannot edit, and it has no taste at all.
Use it. Learn it properly. Charge for it correctly. But if you’re waiting for the magic button, get comfortable, because it is not arriving this year.
Work with us
Coral Film Co is a video production company based on the Sunshine Coast, working with builders, developers, hospitality operators and national brands across South East Queensland. We shoot, edit and finish video in house, and we use AI where it genuinely makes the work better, not because it sounds good in a pitch deck.
If you’ve got a project in mind, or a photo archive you’d like brought to life, get in touch.
About the author. Angus Mead is the founder of Coral Film Co, a Sunshine Coast video production company with more than ten years of trading history. Before starting Coral Film Co he was National Video Chief of Staff at News Corp Australia and Video and Social Editor of Sport at The Daily Telegraph. He works hands on across production and post, and has been testing generative video tools on live client work since they were barely usable.
