tool · beta · on-device AI
Drop a photo, type a word. The AI cuts the subject loose; the text drops in behind them. Magazine-cover, movie-poster, scroll-stopping. Nothing uploads — model and photo both stay in your browser. For real layouts, grab a proper app.
Lining up the layers…
Lining up the layers…
where this trick earns its keep
for portraits
The cleanest way to make a portrait feel designed: drop your first name in a heavy weight, sized to fill the frame, sitting just behind your shoulders. The subject pops forward, the type falls back. Magazine cover energy without the magazine.
for posters
The movie poster look — bold title behind the protagonist — used to need Photoshop and a steady hand with the lasso tool. Now: upload, type the title, hit download. Same effect, no masking.
for instagram
Carousel cover slides win when type fights with the photo. Subject in front, oversized word behind — the eye reads the word as backdrop, not headline, and the photo carries the click. Saves 3× more than text-on-top.
for thumbnails
Shrink your thumbnail to mobile-feed size and only the contrast survives. Text behind the subject keeps the type huge enough to read while the face still draws the eye. Two layers of hierarchy in one frame.
for events
Drop a portrait of the birthday kid, type the age in 600px Black, send it. The number sits behind their head like a backdrop. Way more personal than a stock confetti template, and you didn't pay $4.99 for the printable.
for pets
A cat photo with their name behind them in chunky serif type reads as portrait, not pet meme. Print it, frame it, hang it next to the door. They will not appreciate it. You will.
in case you're wondering
No. The AI subject cutout runs in your browser via WebAssembly — the model downloads once, then your photo never leaves your device. We don't run any image pipeline on a server.
We composite three layers: your original photo on the bottom, your text in the middle, and an AI-cut version of your subject on top. Because the subject sits in front of the text, the text appears to fall behind it.
Photos with one clear, well-lit subject: a portrait, a pet, a single product, a person at full height. Busy group shots, low-contrast subjects, or photos where hair blends into the background can produce a soft cutout edge.
On the first photo of the session we download the segmentation model (~30MB) and warm up WebAssembly. After that, every subsequent photo is much faster. The model is cached for the rest of the session.
Yes — add as many as you like. Each layer can have its own text, font, weight, size, color, and rotation. They all sit behind the subject. Click any layer in the side panel to edit it.
PNG at 2× the canvas resolution. The text bakes into the pixels, and the result has the original photo's aspect ratio. Use it as-is for social, or drop it into a layout app for print.
Not in this tool — it's intentionally minimal. For edge refinement, save the cutout from our Background Remover tool and bring both layers into a real editor, or accept the soft edge as part of the look (it usually reads as styling, not as a flaw).
now what?
next →
Move the whole composition onto a new backdrop.
Opennext →
Reframe the magazine-cover composition as a 1000×1500 pin.
Opennext →
Crop the result to a LinkedIn or X cover.
Opennext →
Dust the cover with Tri-X grain for an editorial finish.
Opennext →
Map the whole image to a two-colour palette to match a brand.
Opennext →
Add a smaller handwritten caption alongside the big behind-subject word.
OpenCut the background out of any photo in one click.
OpenDrop your subject onto paper, gradient, or your own backdrop photo.
OpenNine presets, one photo — vintage to cinematic.
OpenTri-X, Kodak Gold, Lomo — pick a roll, see the grain.
OpenComic dots, risograph red, punk-zine textures.
OpenPhoto to graphite drawing in two clicks.
OpenMap photo to two colors — Spotify, sunset, riso.
OpenMake real cities look like tabletop dioramas.
OpenSunset, lomo red, golden hour — bleed light into corners.
OpenVHS bands, anaglyph 3D, CRT scanlines, chromatic aberration.
OpenConvert any photo to typed characters.
OpenPure invert, cyanotype, blueprint, X-ray, darkroom, sepia negative.
OpenPhoto as tiles with grout-color gaps.
OpenBrightness, contrast, saturation, hue, blur — seven sliders.
OpenAny photo, wrapped in paper with a handwritten caption.
OpenFour photos, B&W mode, handwritten caption.
OpenTwo photos, side-by-side or stacked.
OpenPaper mat for prints, journals, framed walls.
OpenCircle, heart, star, arch — pick a cookie cutter.
OpenGreetings caption, faux stamp, postmark, paper texture.
OpenFloat a photo over paper for portfolios and mockups.
OpenDrop a photo, type a title, pick a layout. 1000×1500 native.
OpenX, LinkedIn, or YouTube cover with title overlay.
OpenHandwritten captions, magazine headers, or a quick scribble.
OpenStick paper-mood tape onto any photo.
OpenLED red, Kodak yellow, camcorder amber.
OpenCorner sign-off or full diagonal tile.
OpenHearts, stars, flowers, smileys, paper tags, HELLO labels.
OpenSquare for IG, 9:16 for stories, 2:3 for Pinterest.
OpenMirror a selfie, un-tilt a scan, turn a sideways shot upright.
OpeniPhone photos that open on everything.
OpenHide faces, plates, addresses — drag a patch over the spot.
OpenRemove GPS, camera info, timestamps before posting.
OpenSmaller JPG, WebP, or PNG — same look, less file.
OpenPull 5 hex codes out of any photo.
OpenPrint-ready QR for menus, invites, journals, vCards.
OpenEyedropper a color in your photo, swap it for another.
OpenFor the real thing →
One word behind one subject is a piece. When you want layered type, gradients, multiple photos, and brand kits — a real design app handles it cleanly.