liminals.space turns your own photo into a backrooms liminal-space image, keeping your actual layout, your real proportions, the corner of the room your brain already knows, instead of generating a random hallway you have no memory of.
4 photos · see 2 rooms free · no account · your photos aren't saved.
Why starting from your photo is the whole point
The feeling that makes liminal spaces hit, that slow stomach-drop of recognition, doesn't come from eerie lighting or empty corridors by themselves. It comes from your brain recognising a specific geometry. The angle of a hallway it filed away twenty years ago. The proportion of a ceiling that was slightly too high. The door at the end that was always slightly wrong.
Generic AI backrooms generators work from text prompts. Type "school hallway" and you get a school hallway, technically correct, emotionally inert. It could be anyone's school. Your brain looks at it and goes: sure, that's a hallway. No recognition, no shiver.
When you start from your own photo, the AI keeps your real layout as the skeleton. The result is your specific room, emptied and stilled, with the fluorescent hum turned up. Your brain looks at it and goes: I know this place. Where is everyone. That gap is where the feeling lives.
liminals.space is the only tool that builds from your photo of a real place you actually went to. Not a prompt, not a stock interior, not a hallway harvested from someone else's memory.
Your photo vs. a generic generator: what's different
Both tools produce eerie, empty rooms. Here's where they diverge:
Generic prompt-to-image: invents a hallway. Looks liminal. Has no connection to your life. The emotional ceiling is "kind of creepy."
liminals.space from your photo: rebuilds a room you already know. The emotional ceiling is "I need to send this to everyone who grew up in that house with me."
The technical difference is that the AI uses your image as a structural anchor, preserving perspective, proportions, and the spatial logic of your specific room, before applying the liminal treatment. A prompt-only model has no anchor; it makes something up.
That's also why people-free shots work best. When the room is the subject rather than the people in it, the AI reads the geometry more clearly and the reconstruction stays truer to the original.
What makes a good source photo
You don't need a good camera. You need a photo where the room is legible. A few things that help:
Show a corner, a hallway axis, or a long wall. Clear geometry gives the AI something to reconstruct from.
No people. People-free shots let the AI focus on the space. If someone's in the frame it'll still work, but emptied rooms read cleaner.
Interiors only. The backrooms aesthetic is defined by its indoors, institutional, no-outside quality. Outdoor photos don't carry the same logic.
Recent photos are fine. The AI handles the era and mood, so you don't need a photo from the 90s to get a 90s feeling.
Upload 4 to 6. The minimum is 4 and the maximum is 6. More angles from the same space give you more rooms, each with a slightly different vantage on the same memory.
What you get
After uploading, the AI processes your photos and rebuilds each one as its backrooms version. The first 2 rooms render free, in lower resolution, enough to see whether it worked. Then:
Full resolution rooms: every image in full quality.
3D step-inside viewer: explore one of your rooms in three dimensions, inside your browser, no download or install required.
VHS-style nostalgia video: a short film version of your space, with the scan lines and ambient hum that make it feel like a found tape.
Keepsake download: a PDF and zip of your rooms to keep, share, or print.
All of this comes with a one-time payment, no subscription, no account. An optional higher tier also removes the watermark and ambient music if you want clean images for posting. Your photos are used once to generate the output and then deleted; nothing is stored, and no one else ever sees your rooms.
Which room to start with
Some places carry more of that specific geometry. Start wherever the memory is strongest:
Recognition is the emotional core. Generic generators invent a hallway you have no memory of, eerie but impersonal. When liminals.space starts from your photo, it keeps your real layout and proportions, so the result feels like your specific memory surfacing at 3am, not a stranger's.
What kind of photo works best?
People-free interior shots with clear geometry: a corner, a hallway axis, a long wall. Phone snapshots are fine. Recent photos work just as well as old ones; the AI handles the era and mood.
What do I get when I unlock the full set?
Full-resolution rooms, a 3D step-inside viewer in your browser, a VHS-style nostalgia video, and a downloadable keepsake (PDF + zip). A higher tier removes the watermark and music for clean posting. One-time payment, no subscription, no account.