Case study · Olarm ONE launch

From a ten-minute shoot to a launch hero

The Olarm ONE range needed a launch image. I shot the real boxes on an office table, used ChatGPT to build the set, light and story around them, then edited every product detail the model got wrong in Photoshop to match the real packaging.

10 min

Reference shoot, real boxes

ChatGPT generation kept as the base

Product errors caught and corrected

10,000 px

Square master, cropped per format

Objective

A premium launch image for the Olarm ONE wireless range, true to the real packaging, with room for a headline.

My role

Shoot, art direction, AI generation, Photoshop retouching and the final social layout.

Tools

Camera · ChatGPT image generation · Photoshop · Figma

Output

A 10,000 px square master and the LinkedIn launch post.

01 · Shoot

Ten minutes, real boxes

No studio, no stylist: the real packs on an office table, shot from a few angles. These frames were never meant to be the image. They’re the reference, so the model starts from the actual boxes, colours and layout.

02 · Generate

ChatGPT for the set, light and story

From the reference, ChatGPT built what the office couldn’t: a dark seamless set, controlled light, and a hand lifting the HUB lid. The mood was right in one pass. The product wasn’t.

What it got right

Set, light and depth that read as a studio shot
A human moment: the HUB being unboxed
A layout that holds the whole range

What it got wrong

Invented two products that don’t exist
Misspelled KEY and garbled the small print
Washed the brand navy out to slate
Left the HUB dark and lifeless

03 · Fix

Photoshop for the truth

AI handles atmosphere. Customers check the box. I kept ChatGPT’s set and light, then edited each pack to match the real packaging. Drag across to compare.

ChatGPTFinal

Detail before and after: an invented product, replaced. Left, the ChatGPT output; right, the retouched final. — 01An invented product, replacedChatGPT made up an “App” box. I edited it to match the real 90XS outdoor detector: pack artwork, copy and the product itself.

Detail before and after: the wrong device, corrected. Left, the ChatGPT output; right, the retouched final. — 02The wrong device, correctedThe “Sight” box showed a remote that isn’t in the range. It became the real ENTRY sensor, with the right product and copy.

Detail before and after: the hub, switched on. Left, the ChatGPT output; right, the retouched final. — 03The HUB, switched onThe model left the HUB dark. Status lights make it read as a live device, and give the hand reaching in a reason to be there.

Detail before and after: name, colour and artwork restored. Left, the ChatGPT output; right, the retouched final. — 04Name, colour and artwork restored“REY” became KEY, the faded slate went back to brand navy, and the real six-button remote replaced a lookalike.

04 · Ship

Room for the message

I extended the canvas to a 10,000 px square and let the set fall to black, so the headline gets its own space and one master crops cleanly to every format. First stop: the LinkedIn launch post.

Square master with the canvas extended into dark space above the products — Square master · canvas extended for copy

LinkedIn launch post as published: “ONE less thing to wire”, introducing Olarm ONE, wireless security for homes and businesses. Wireless, up to 2 km range, easy to install. — LinkedIn launch post, as published · 1080 × 1350 · See it on LinkedIn ↗

Start → finish

Start: phone-quality reference frame on an office table — Office table, ten minutes

Finish: the retouched launch image — Launch hero

The takeaway

AI did the set, the light and the story. The parts a customer checks (names, labels, colours, the light on the HUB) took a designer.

Back to AI image production