Black Friday SaleEnds in0 00:00:00
Exploring Backroom
A ChatGPT image prompt written and tested by Visual AI Club.
Copy the full prompt with the button below and paste it into ChatGPT.
The guide further down explains each part and what to change.Prompt guide
This prompt drops two people and their two cats into an endless abandoned indoor pool complex, a liminal Backrooms-style space, with a giant, slightly grimy white swan floating at its center. The group stays small in the lower part of the frame while tiled halls, columns and turquoise water fill most of the picture.
By the Visual AI Club team · How these guides are written
Describe each cat in words
Add a coat line such as grey tabby with white paws, because a curled cat with a hidden face lets the model invent a new breed.
Rewrite every count for your group
For one person and a dog, edit each mention of two people and two cats, and drop the beard line if nobody has one.
Confirm the hall dwarfs the group
Columns, tiles and the swan should fill most of the frame, with all four characters small in the lower part.
Do this
The characters are too big
Ask for a wider view and place columns both in front of and behind the small group.
An extra animal appeared
Search the prompt for a leftover mention of two cats and change it to match your photo.
A cat came out as a different breed
Use a source photo with that cat's face visible and spell out its fur pattern and color.
The space feels crowded
Cut back to the few allowed props: a red balloon, a LEVEL 0 mark, empty chairs and an old waterslide.
If you want
replace the swan everywhere with a giant faded rubber duckIf you want
a few more flickering ceiling panels over the far corridorOne idea repeats across several blocks: the people must not dominate. The architecture and swan should take up roughly 75 to 80 percent of the frame, with the group in the lower left or center. The swan is several times taller than the people, and corridors run off into darkness.
That is what makes the picture work. A giant swan beside big, close-up faces just looks like a collage; small figures among repeating columns, balconies and reflections make the space feel impossibly large.
The identity lock is unusually detailed for animals, covering fur pattern, color, facial features and proportions for both cats, plus the relationship between all four characters. The characters block then assigns roles: a bearded man in front with a flashlight, a woman with a lantern, and both cats walking ahead in small harnesses.
If your photo shows a different group, rewrite those lines. For one person and a dog, change the counts everywhere they appear, because the prompt mentions two people and two cats many times and also forbids removing either cat. A leftover mention is the usual reason an extra animal shows up. Remove the beard reference if nobody has one.
LEVEL 0 mark, empty white plastic chairs and an old waterslide.Cats are the hardest part. Use a photo where both cats are clearly visible with their faces showing; a curled-up cat with its face hidden gives the model nothing to match, and it may invent a different breed. It also helps to describe each coat in words, such as grey tabby with white paws.
The people's faces should be reasonably large in the source, because they end up small in the result. Small faces in, small faces out tends to lose resemblance.
Yes, as long as you keep the single central element idea. A giant rubber duck or a floating paper boat works well. Change it everywhere the swan is mentioned, including the final image section.
Ask for a wider view and repeat that the four characters should fill only a small part of the lower frame. Mentioning columns both in front of and behind the group helps the eye read the scale.
Next prompt · ChatGPTGTA VI Artwork PromptOpenLearn AI Academy
Copy prompts that work, then learn why they work. The Learn AI Academy teaches AI image and AI video creation step by step: writing prompts, light and camera, consistent characters, product shots and directing a short film with Seedance.
Open the Learn AI Academy