Before and after, all six
My own files, one word each, nothing retouched afterwards. The interesting part is not that it works. It is where it quietly changes things.
- PREMIUMSHOWCASE
The one that actually landed
A phone snapshot of a mug lying on a table became a styled product shot: plant, stacked books, a poster in the background, soft daylight. This is the result that stuck with me.
Before 
After Uninvited extra: it printed the word PREMIUMSHOWCASE into the image as a headline, plus the tagline Simple things. Big joy. I only typed one word.
- 360VIEW
Ten angles, one glitch that would not go
Same mug, different word. Ten angles from a single photo, including the 90 degree top view where only the handle is visible. The level of detail is real.
Before 
After Look at 45 degrees and 360 degrees: the Game Boy character has three arms. On the original photo it has two. The error repeats in every angle where the figure is visible, and ChatGPT's own eraser tool did not remove it.
- XRAY
Pure fun, zero use
I typed X-ray full skeleton on a photo of my dog. The fur stays fur, the harness stays solid, the skeleton sits underneath. Technically clean, practically useless for me.
Before 
After - ADCREATIVE
Polished, but not a transformation
My own book cover. The result is cleaner, better lit and straightened up, which says more about my original photo than about the prompt. It is my cover, sharper. Not the ad creative those viral posts promise.
Before 
After I actually uploaded two photos of the book. The second one was visibly ignored in the result, which is worth knowing if you plan to feed a prompt several angles.
- HDREAL
Clearer photo, invented tattoo
A photo from the Dresden Marathon. The blur at the edges is gone and the image is noticeably clearer. Then I looked at my forearm.
Before 
After The tattoo reads Be strong, believe. In the result it is shaped similarly but the letters are nonsense. Most people would scroll past it. That is exactly the problem.
- TOGETHER
Two photos, years apart, two mistakes
Two pictures of me with different hair colour, merged into one. Both faces are clearly me, the book in my hand faces the right way. Technically this is the hardest thing in the whole test, and it mostly worked.
Input 
Input 
Result Two errors in one image: my Garmin became a square smartwatch, and my hoodie, which reads Strongminded Productions, came back as Long-minded Production. Neither is subtle once you know where to look.