Services
Case study · 04 Content Creation

Six thousand assets, judged by people paying for them.

The stock libraries are where the generative work gets marked by strangers with a budget. No curation, no portfolio framing, no explaining the thinking behind it. An image is either licensed or it is not, and the dashboard does not care how much you liked it.

Client
Open market · Adobe Stock, Magnific
Year
Ongoing
Engagement
Generative imagery · prompt engineering · finishing
Role
Contributor
6,109 Assets in the Adobe Stock portfolio Visible on the contributor dashboard. Not a figure I can edit or a selection I curated.
1,696 Lifetime downloads · Adobe Stock Each one is somebody choosing this image over every other result on the page.
4,356 Total downloads · Magnific, level 3 A second platform with a different audience, which is what stops one library flattering the numbers.
01 — Why this is the honest test

A portfolio is a curated argument.

Every portfolio on the internet, including mine, is a selection. You show the ten pieces that worked and quietly retire everything else. That is not dishonest, but it does mean a portfolio cannot tell you what somebody produces on an average Tuesday.

Stock does not allow the edit. Every asset goes up. The market licenses some and ignores the rest, and the ratio is public on a dashboard I do not control. Across six thousand assets that average says something a showreel structurally cannot: whether the work is actually useful to somebody who has a job to finish today.

It is also where the generative stack gets stress-tested. Prompt engineering, model selection, upscaling and colour finishing all get judged by whether a stranger spends money, which is a considerably harsher critic than a client who already likes you.

02 — What the market taught

Four things, repeatedly.

01 Direct it

A prompt is a brief. Vague in, generic out — and generic does not license, because generic is what every other contributor also uploaded.

02 Generate wide

Volume is cheap at this stage and expensive later. Choosing from forty is a different job from rescuing one.

03 Finish properly

Upscale, clean the artefacts, grade. Most rejected stock fails on finishing rather than on the idea behind it.

04 Let the market mark it

Downloads are the only review that counts. What sells decides what gets made next, not what was satisfying to make.

03 — What a client gets

The same process, pointed at your brand.

Generative imagery sets
Ten art-directed brand images, upscaled, finished and commercially licensed. Priced against stock licensing plus a design half-day.
Thumbnails and ad creative
Built for the specific platform they will be judged on, not resized afterwards and hoped for the best.
On-site photo and video
Half or full days. Roughly four to six usable clips and fifteen stills from a half-day, edited and delivered.
Graphics packs
Ten branded assets from a single brief, with a revision round on each, so the whole set stays consistent.
04 — Disclosure

You decide what to say, not me.

Generative imagery carries real commercial questions, and the people selling it hardest are usually the ones least interested in raising them.

There are three worth knowing about: what a model was trained on, what rights actually come with the output, and whether your customers would mind knowing a campaign image was generated rather than photographed.

How I handle it. I tell you which assets in a set are generated and which are photographed, every time, in writing. Some brands want that line drawn publicly and some genuinely do not care — both are reasonable positions, and it is your call to make rather than mine to quietly avoid. Where a use case is legally uncomfortable, such as recognisable faces or anything implying endorsement, I will say so and we shoot it instead.