object ⇄ fill
Remove the object, fill the background, and keep the object
Every cleanup tool can erase an object. This one lifts it onto its own layer and fills the background behind it, so "remove" becomes "move", and you can always change your mind.
Lift an object from a photoErasing is a one-way door
Cleanup-style tools inpaint over the object and hand you a flattened result: the object is gone, forever, and if the fill has an artifact you start over. The layered version of the same operation keeps everything: object on one layer, reconstructed background on another. Slide the object aside, scale it, duplicate it, or hide it: the "removal" is just a visibility toggle.
How the fill works
Select the object (one AI click), Lift selection, then Fill behind. The editor pre-fills the hole with surrounding texture and the server model refines it into a photographic patch: an approach we adopted after testing because it avoids the classic inpainting failure of hallucinating new objects into large holes. Filled regions are auto-prefixed AI_Fill_ in the layer names.
The result, unretouched
This is real product output from our validation runs: the subject-shaped hole in the background, filled. Small object-sized fills are typically seamless; very large fills (an entire person) are honest-but-soft, and the object layer usually covers them anyway.
A file, not a screenshot
Export the whole thing as a layered PSD — object layer, filled background, masks intact, and keep editing in Photoshop, Photopea, Affinity, GIMP, or Krita. Cleanup tools give you a picture; this gives you the working document.
Frequently asked
How is this different from Cleanup.pictures or Magic Eraser?
Those erase and flatten. This lifts the object to a layer and fills behind it — same visual result if you hide the layer, but reversible and editable.
What does it cost?
Selecting and lifting are free (in-browser). The AI fill uses the server: 2 credits per fill. New accounts get 10 free credits.
How good are the fills?
Object-sized holes fill seamlessly — see the samples page. Very large holes (a full person against a detailed scene) come out soft rather than invented; we deliberately tuned the model so it never hallucinates new objects into your photo.
Related
Material separation for Live2D, with AI doing the boring halfYour product photos, as layered files you ownSplit a background into parallax planes for your 2D gameHow PSD layer writing actually worksCanva can't make layers from an image. Here's the workaround.Get image layers into PowerPoint: the clean wayMake a depth-effect wallpaper from your own imageSeparate the subject from the background, and keep bothPull the text out of an image — as text you can actually editMost "image to PSD" converters are fake. Here is the 30-second test.Color separation for screen printing, without the $500 softwareTurn a flat image into a PSD with real layersSplit any image into separate, editable layersSplit an image into layers — by object, or by colorAn AI layer separator that hands you a real fileTurn any image into a layered .ora for GIMP and KritaOne photo in. The layered file you actually need, out.