Fur Real. A franchise ad campaign made with AI, while the tools were still learning to look real.
Art director Adrian Elton needed photographic key visuals for a Dash Dog Wash campaign. Commissioned photography was beyond the budget and stock was too limiting, so he brought me in to create the campaign's people and dogs with AI.
Five people. Five dogs. One look of disbelief.
Each poster pairs a would-be owner with their dog, caught in the same moment of happy surprise. Five personas, one consistent look.
As both graphic designer and art director, I had been playing around with AI generated imagery from the earliest days of Dall-E back in mid 2022. Despite the rudimentary nature of those first crude images I could immediately see its potential for creating fully fledged advertising imagery.
Towards the end of 2022 I got a brief for an ad campaign for a dog wash franchise called ‘Dash Dog Wash’. Once the 'fur real' campaign concept had been selected and the discussions turned to production, I added a suggestion to the mix that I’d never been able to suggest before; AI. With commissioned photography falling far beyond the production budget, and stock imagery being far too limiting, it was clear to me that AI would give us the best of all scenarios.
As my AI skill levels were still very basic, I knew we had to wheel in the big guns to produce imagery with the requisite level of complexity.
I had already been in conversation with various AI artists whose work had wowed me, including the incredibly talented Brian Sykes. In particular, I'd seen an image that he'd generated of a greyhound that had an unbelievable level of fine 'photographic' detail. That pooch in mind, I reached out to Brian to see if he could bring that level of realism to the task of creating the key visuals for our campaign.
As became apparent once the process began in earnest though, it's one thing to create fantastical AI images that are an end unto themselves, and quite another to create a complex and coordinated suite of advertising images that have campaign level consistency.
Across the trajectory of the project we discovered that in order to get the necessary facial expressions of surprised and delighted awe, the results were far stronger when prompting the humans and hounds separately. So even though there was more overall realism by prompting them together, it was impossible to simultaneously get their expressions right, while also getting their poses sufficiently consistent.
The next big problem was that MidJourney 4.0 still gave everything a Marvel 3D like illustrative quality that was incredibly visually alluring on the one level, but was still fundamentally at odds with the photographic aesthetic that was the brief requirement from day one.
I'm not sure how we would have overcome this otherwise, but the game-changing moment was when MidJourney 5.0 became available, mid-project, allowing us to re-run and refine all of the prompts. This had the effect of transforming the images into something that for the first time felt truly photographic, relative to what had come before.
Along the way Brian very generously and patiently showed me the ropes as I had an irrepressible eagerness to get up to speed with MidJourney having seen up close and personal what he was capable of producing.
This turned out to be incredibly prescient given that the project had more than eclipsed Brian's availability in the midst of his hectic schedule. Indeed, with Brian's patient stewardship, I was able to complete the project while also adding the oh-so-contemporary skill set of 'AI Whispering' to my CV.
All up, the MidJourney was just as valuable as the destination.
Casting a campaign, one face at a time.
First, was referencing the campaign plan, and developing images that fit the defined profiles. Adrian had developed 5 personas with a description of each we were looking to capture. With each personality, there was also a few dog breeds for consideration to be included with the person to complete the ensemble.
The challenge was working across 5 flavors of humanity - male and female, with diversity of age, backgrounds, styling of dress, expressions of humanity... while at the same time obtaining a consistent look and feel for all the people and pets. We literally ran through hundreds of possibilities to land on the right individuals.
Now think of how you might describe that with a prompt to get a consistent look out of your characters - while retaining individuality.
Also, while you are at it - position the heads the same, make sure the dog is positioned in the same spot each time (and we want their face in a derpy expression). Everyone's eyes need to look the same way. We need outfits to match personality....
Did I mention... you have a limit on effective characters - or the actual number of words you can use in the prompt to accomplish everything you want... considering male/female, ethnicity, age, outfit, hairstyle, special features, dog, expressions, lighting... That is where this got complicated quickly, but I developed a process to solve the issues and address the workflow. This, as Adrian mentioned, was before 5.2, SDXL or Photoshop Beta.
![The [Hu]man Element](/wp-content/uploads/2026/10/bs-hu-element.png)