A whole product shoot, from one paragraph
AI product photography that ships: eight images on a warm seamless, label readable at thumbnail size, from a brief that took under a minute to write.
What was the brief?
'Eight studio images of the refill pouch on a warm seamless ground, editorial lighting, label legible at thumbnail size.'
That is the whole thing. The product itself was never described, because the workspace already knew it — the pouch was in the cast, with reference frames attached and approved weeks earlier. That single fact is most of why the brief could be one sentence.
Why does 'legible at thumbnail size' matter more than resolution?
Because that is where the image will actually be seen. A product shot lives in a grid, on a phone, at about the size of a postage stamp, competing with eleven others.
Specifying the condition the image has to survive is a far better brief than specifying the pixels. Resolution is easy and nearly always irrelevant to whether the shot works.
What happened before anything was spent?
Two refusals, both immediate, and neither of them a surprise once read.
The first draft asked for a competitor's bottle in frame for scale, which is a third-party mark on a product. The second added a hand model with no consent recorded. Each refusal named the reason in one sentence and offered the routes still open, and the brief was never lost — you edit and carry on rather than start again.
What did the run itself look like?
The price was on screen before it started. It was approved, and it worked.
Eight images came back on the same ground, in the same light, with the same pouch — the sameness being the entire point, and the thing that is hardest to get from a tool that has never heard of your pouch before.
What was flagged?
One frame. The label curved further than it should where it wrapped the seam. Flagged, not fixed.
It is the kind of thing a person scanning eight thumbnails would miss on a Tuesday afternoon and a client would find on Thursday morning. Catching it is worth more than the frame it cost.
Why flag it instead of quietly fixing it?
Because a quiet fix is the system renegotiating your brief on its own.
The label comes back legible and slightly wrong. The product comes back consistent and slightly generic. Runs get cleaner and the work gets worse, and the metrics all improve while you start needing a second round on sets that passed everything. So it stops and asks, and you spend four seconds deciding.
How did it end?
Seven kept, one rejected and rerun. The whole thing sat inside a coffee break.
The finished set went to the client on a link, and the record says who approved what. Nothing about this run is remarkable, which is the point — it is what a Tuesday looks like when the boring parts have stopped being your job.
What would have made this harder?
Not having the product in the cast. Everything good about this run traces back to a piece of setup somebody did weeks earlier: uploading real photographs of the pouch, naming it, and approving it once.
Without that, the brief is three paragraphs, the pouch is subtly different in image six, and you find out at review.
Does this replace a photographer?
For eight pack shots on a seamless, largely yes, and that was never the interesting part of a photographer's job.
For the shoot that establishes what your product looks like in the first place — the one that produces the references everything else is built on — no. That shoot got more valuable, not less, because now it feeds everything.
What makes a product brief succeed or fail?
The product being known, and the condition being named.
Known means real photographs are in the library and approved, so nothing is invented. Named means you said what the image has to survive — a thumbnail, a dark background, a crop to square — rather than what it should look like.
Briefs fail on the first far more often than the second, and the failure does not look like failure. It looks like eight nice images of something not quite your product.
What should you upload before your first run?
Three to five honest photographs of the product from different angles, unretouched, including whichever angle shows the label or the detail people recognise it by.
Phone photographs in ordinary light are genuinely fine here and often better than a catalogue render, because they contain the real object rather than an idealised one.
Then approve it once, and never describe it again.
How does this compare with a studio day?
A studio day is still the right answer for the shot that defines the product — and it becomes more valuable, not less, because the images it produces are what everything downstream is built from.
What changes is the long tail: the seasonal variant, the extra crop, the third background someone asked for on Thursday. That work used to require another day or a compromise, and now it does not.