The 2.5D parallax effect: how it works and how to make one from a single photo

A 2.5D parallax effect makes a flat photo look three-dimensional by splitting it into layers at different depths and moving them at different speeds as the camera shifts. The near layer moves more than the far one, and your brain reads that gap as depth. You can build the layers by hand, or let a depth model build them from one photo in a few seconds.

What a 2.5D parallax effect is

Think of a pop-up book. Every piece of card is flat, but the pieces sit at different distances from you, so when you tilt the book they slide past each other. That's the whole trick. A 2.5D image is a stack of flat cut-outs (or one surface pushed in and out by a depth map) viewed through a camera that moves a little.

The “half” in 2.5D is honest labelling. You get depth, but only from the one angle the photo was taken from. Swing the camera too far and you'd expect to see the side of the person's face, which the photo never recorded. So every good 2.5D effect keeps the move small. Gifsy stops at about 20 degrees each way.

Why it reads as 3D

Motion parallax is one of the strongest depth cues you have. Look out of a train window: the fence posts blur past, the hills barely move. Nobody taught you that nearer means faster, but you read it instantly. A 2.5D effect fakes exactly that relationship inside one picture.

The second cue is occlusion. As the subject slides sideways, a sliver of background that was hidden behind it comes into view. If that sliver is empty or smeared, the illusion breaks on the spot. That's why the background behind the subject has to be filled in, whichever route you take. Most of the work in a convincing 2.5D image is that fill, not the motion.

The manual route: cut the layers yourself

This is the classic workflow in most tutorials, and it still gives the most control:

1. Select the subject in Photoshop and move it to its own layer.
2. Fill the hole it leaves in the background (Content-Aware Fill gets you most of the way; the rest is clone-stamping).
3. Split further if the scene has a clear middle ground: a table, a street, a hill.
4. Bring the layers into After Effects, space them out in Z as 3D layers, and animate a camera.
5. Render a video.

It works well for a hero shot you'll use for months. It's slow for anything else, mostly because of the background fill, and the result is a fixed video. Every layer is also perfectly flat, so a face or a curved product reads like a cardboard stand-up unless you add displacement on top.

The one-photo route: let a depth model do it

A monocular depth model looks at one photo and estimates how far away every pixel is. Gifsy uses Depth Anything V2 for this; you can see its raw output with the free depth map generator. Bright is near, dark is far.

That map replaces hand-cut layers with a continuous surface. Instead of three flat cards you get a mesh that bulges where the nose is and recedes where the wall is. Gifsy then does four more things:

• Lifts the subject. A second model cuts the subject out, so it becomes its own plane in front of the background.
• Fills the gap. When you publish, an inpainting model paints the background behind the subject once, so the reveal shows scenery rather than a smear.
• Moves each point by its depth. Every vertex shifts by pointer × strength × (depth − pivot). Points in front of the pivot move with the cursor, points behind it move against it.
• Eases the motion. One smoothed driver feeds every mode, so switching from a slow drift to following the pointer never snaps.

The full pipeline, with the numbers, is in how 3D photos work.

Video or interactive: pick the output first

Most 2.5D tools render the camera move once and give you a file. That's the right call for Instagram, TikTok or an email. On a website you have a second option, a scene the visitor drives with their own pointer. Decide which you need before you pick a tool.

Gifsy does both: the same scene exports as a GIF or WebM, and publishes as an embed. The comparison of every tool we looked at is on the 3D photo makers page.

Make a 2.5D parallax effect online from a single photo

1. Pick one photo. One clear subject, some space around it, 640 px or more on the long edge.
2. Generate the depth. Upload it at gifsy.fun/create. A depth model estimates distance per pixel and a second model lifts the subject.
3. Tune the move. Choose how far it moves and how it's driven: pointer, drag, scroll or a slow drift.
4. Export or embed. Download a GIF, WebM or PNG, or publish for a share link and an iframe for any site.

3D on Gifsy needs a free account, and making scenes is free and unlimited. Your photo stays in the browser while the scene is made; the finished scene is uploaded only when you publish. Open the 3D parallax photo maker.

Where 2.5D breaks

Every 2.5D image fails in the same few places, whatever made it:

• Subjects touching the frame edge. Nothing exists past the edge, so the side of the subject tears as it moves.
• Glass, mirrors and water. The depth model sees the reflection, not the surface.
• Fine hair against a busy background. The cut-out edge gets soft and the hair drags the background with it.
• Flat, low-contrast scenes. With no clear near and far, the whole picture warps like a sheet instead of separating.

The photos that work best, with a quick test, are in best photos for a 3D effect.

FAQ

Make a 2.5D parallax photo. Related: a 3D image effect on mouse movement and interactive 3D photos for websites.