A feature on the budget that no longer writes the brief
Anyone can generate an image. We art-direct one.
AI can now produce a striking image or a moving sequence on demand — and, just as easily, a flood of generic, uncanny, off-brand and legally-dubious visuals. The generation was never the hard part; the direction is. The value lies in holding a single, consistent brand look across an entire campaign, in the taste to reject the subtly-wrong frame, in handling rights and safety properly, and in knowing which shots AI should make and which it shouldn't. We use these tools to produce a campaign's worth of imagery for the budget of a single shoot — directed, not merely prompted.
A creative director at a wall of AI-generated visuals, selecting the on-brand frames and rejecting the uncanny ones, a brand look-book pinned alongside. Art direction applied to generation. Square aspect ratio, dark studio, focused light, solar-yellow glow, a sense of taste and selection.
Generation is cheap; direction is the craft — one brand look held across a campaign, the uncanny frames cut.
AI has collapsed the cost of visuals almost to nothing. What once required a photographer, a studio, a crew, a location and a full day now takes a prompt and a minute — and that shift is genuinely transformative for any brand that lives on images. It is also a minefield. The very same tools that produce a beautiful frame will, with equal confidence, hand you uncanny hands, an off-brand mood, an accidental near-copy of someone else's protected work, and a flat sameness that screams "made by AI." The cost of making an image fell to zero; the cost of making the right image did not.
The hardest problem, and the one that separates real work from novelty, is consistency. A brand campaign is not one striking image; it is a family of them that share a look — the same character, the same lighting language, the same world, across a dozen frames and formats. Generative tools are brilliant at producing one arresting image and notoriously bad at producing the next one to match it. Holding a coherent identity across a whole set — a recurring face, a consistent style, a unified palette — is where amateur output falls apart and where genuine craft and the right techniques earn their keep.
The other half of the craft is editorial: knowing which frames to keep. Generation produces volume, and most of that volume is subtly wrong — a hand with too many fingers, an expression that lands in the uncanny valley, a detail that breaks the brand or, worse, that echoes a real artist's protected style or a recognisable person's face. An experienced eye does the unglamorous, essential work of culling: rejecting the eighty per cent that fails, fixing the salvageable, and approving only the frames that are genuinely on-brand, genuinely usable and genuinely safe to publish. The taste to say no is the entire job.
In this feature
Six things serious AI image & video gets right.
Brand consistency
A campaign is a family of images that share a look, not one lucky frame. We hold a consistent character, palette and style across the whole set — the hard part of generative work, and the part that separates a campaign from a novelty.
Direction, not prompting
Typing a prompt is not art direction. We bring a concept, a brand brief and a trained eye — composition, mood, intent — and use the tools to realise it, so the output expresses a deliberate vision rather than the model's defaults.
Killing the uncanny
Most generated frames are subtly wrong — odd hands, dead eyes, a detail that breaks the spell. We cull ruthlessly and fix what's salvageable, approving only frames that pass as intentional and real, never the eerie near-misses that betray the source.
Rights & safety
A generated image can unknowingly echo a protected style or a real person's face. We work to keep output original and safe to publish — mindful of likeness, IP and the brand's exposure — because a striking image you can't legally use is worthless.
Variation, one budget
Once a look is locked, producing more — new formats, sizes, seasonal variants, fresh angles — costs a fraction of a reshoot. We exploit that to give a campaign reach and longevity that a single traditional shoot never could.
The right tool per shot
AI isn't always the answer. Some shots still want a camera, a real product or a real face. We choose generation where it genuinely serves the work and traditional production where it doesn't — matching the method to the frame, not forcing it.
The budget no longer
writes the brief.
For most of advertising history, the budget wrote the brief. You wanted a model on a beach in three countries, a product shot from twelve angles, a seasonal refresh every quarter — and the cost of crews, studios, travel and time decided how much of that you could actually have. Brands trimmed their ambition to fit the shoot. AI quietly removed that constraint. The creative ceiling is no longer set by what you can afford to photograph, but by what you can imagine and direct — which is a far better problem to have, and a far more demanding one, because now the only limit is the quality of the direction.
The defining challenge of doing this seriously is consistency, and it is where most attempts quietly fail. Anyone can coax one beautiful image out of a generator; the difficulty is the second, the tenth and the hundredth that have to belong to the same world — the same recurring character with the same face, the same lighting, the same palette and feel, across stills and motion and every format a campaign needs. This is exactly the discipline behind a believable synthetic brand presence: building and maintaining a consistent visual identity that holds up shot after shot. Without it you have a scrapbook of unrelated images; with it you have a campaign.
Then there is the matter every responsible brand has to take seriously: rights and safety. A generative model has learned from a vast pool of existing images, and left unguided it can produce output that echoes a living artist's distinctive style too closely, or conjures a face indistinguishable from a real, identifiable person. Publishing either is a legal and reputational risk a brand cannot afford. We work deliberately to keep generated output original and safe — steering away from protected styles and real likenesses, and treating publishability, not just beauty, as a condition of approval. A gorgeous frame you can't legally use is not an asset; it's a liability with good lighting.
The serious version of AI image and video is a production process, not a prompt box. We start with a concept and a brand brief, establish the consistent look and the recurring elements, generate at volume, then cull, refine and clear for rights — exactly the workflow of a real shoot, with generation in place of the camera and an art director still firmly in charge. And we are honest about the tool's limits: where a shot genuinely needs a real product, a real place or a real face, we say so and shoot it. The method serves the work; the work is never bent to fit the method.
A campaign contact sheet of AI-generated frames sharing one consistent look — the same character and palette across many shots — with off-brand frames struck through. Consistency held across a set. Wide cinematic 21:9 crop, dark elegant studio backdrop, solar-yellow accent, a sense of curation and craft.
A whole campaign, one look — consistency held across every frame and format, the uncanny culled, the rights cleared.
Five questions we ask before generating a single frame.
The cost of making an image fell to zero. The cost of making the right image did not — it just moved entirely into the direction.
AI image and video connects across the wider estate. The same human-in-the-loop discipline runs through AI content generation, which handles the words to its pictures; the consistency techniques are exactly those behind AI avatars and synthetic influencers in our content work, where a single recurring face must hold across endless content; AI voice can give that face a way to speak; and AI strategy decides, first of all, whether generated visuals are the right answer for a given brand at all. Generation is a capability; direction is what makes it a campaign.
We produce AI visuals the way a studio runs a shoot — concept first, a consistent look locked, relentless culling, rights cleared, and a real art director in charge throughout. The brief is a campaign that looks directed and owns its rights — not a pile of impressive, unusable, off-brand frames. That combination of consistency, taste and responsibility is exactly why this is art direction rather than generation: the tool made the images cheap, and made the judgement about them more valuable than ever.
A campaign laid out across formats — hero, social, seasonal cuts — all sharing one consistent AI-generated look, a director approving the set. The moment generation becomes a real campaign. Contemporary, shallow depth of field, dark desk, solar-yellow screen glow.
One look, every format — a complete campaign produced for a fraction of a single shoot.
Representative scenario · not a named client engagement
A consumer brand needed a full campaign on a single shoot's budget — until art-directed AI gave it the whole thing for a fraction.
The brand had real ambition and a modest budget — the familiar squeeze. They wanted a proper campaign: a strong hero concept, a family of supporting images, social variants in multiple aspect ratios, and seasonal refreshes through the year. Conventionally, that meant a shoot, a crew and a cost that their budget could only stretch to once, forcing them to cut the ambition to fit. The brief was being written by the invoice, not the idea.
We ran it as a directed AI production. We developed the concept and a precise brand look, then locked the consistent elements — the recurring style, palette and treatment — that would hold the whole campaign together. From there we generated at volume and culled hard, keeping only the frames that were on-brand, free of uncanny tells and safe on rights, retouching the salvageable ones. The result was a complete, coherent campaign across every format the brand needed — produced for roughly a tenth of a traditional shoot, and refreshable on demand.
A full, consistent, multi-format campaign — for roughly a tenth of a single conventional shoot, and refreshable any time.
The change went beyond the saving. Because variations now cost so little, the brand could test more creative, tailor visuals to each channel, and refresh seasonally without a new production each time — reach and longevity a one-off shoot could never give. And because the work was directed and rights-cleared rather than just generated, it looked deliberate and was safe to run everywhere. The budget stopped writing the brief, and the brand finally got the campaign it had imagined rather than the one it could afford to photograph.
We had the budget for one shoot and the ambition for ten. Revolutionize art-directed the whole campaign with AI — consistent, on-brand, and cleared for use — for a fraction of what a single shoot would have cost. It actually looks directed, not generated. The difference, they told us, was that someone was still saying no.
What serious AI image & video work actually involves.
AI visual production is an investment in getting a campaign's worth of imagery without a campaign's budget, and it is scoped to the concept and the output. Engagements range from a single art-directed concept or hero set to a full multi-format campaign with a consistent identity and ongoing refreshes, and the scope rises with the number of looks, the difficulty of holding consistency, and whether motion as well as stills is involved.
The two largest variables are consistency and motion. A handful of standalone stills is a contained piece of work; a campaign that must hold one recurring look across many frames and formats — or moving image, which is far harder to keep coherent — is a larger one. We always include the art direction and culling in the cost, because that judgement, not the raw generation, is what turns output into usable, on-brand, publishable work.
Every engagement includes the full discipline: concept and art direction, a locked consistent look, generation at volume, ruthless culling and retouching, a rights-and-likeness check, and delivery across the formats you need — with the option of ongoing refreshes, since variations cost a fraction of the original once the look is set.
We scope every project to the work rather than to a price list — which is why we don't publish rate cards. Every engagement begins with a free 30-minute scoping conversation, and we will tell you honestly where AI visuals will serve the brand and where a real camera still wins. We would rather direct fewer frames properly than generate a thousand you can't safely or proudly use.
When you're ready
Stop letting the budget write the brief.
Tell us the campaign you'd shoot if budget were no object — the hero, the variants, the formats. We'll respond within 24 hours with an honest read on how much of it AI could art-direct into existence for a fraction of the cost, and which shots still want a camera.
Begin the conversation →