Long Weekend Labs

Parallax Studio

Turn a still image into a looping 2.5D parallax animation. Paint depth by hand, or let it generate depth for you. Every model runs on your own machine, with no accounts, API keys or uploads.

Checking latest release...
Runs on
Windows, Linux, macOS
AI
Offline, nothing uploaded
Project file
.parlx
A still landscape looping as a 2.5D parallax animation
One still image, no layers and no masking. Auto Depth worked out the hills and the tree line; this is the loop it exports.
The Parallax Studio window

What it does

One focused workflow instead of a giant editing suite.

Auto Depth

Depth Anything V2 separates a subject from its background, then the map is feathered where it would otherwise tear: a hard outline does not stretch under displacement, it folds back over itself.

Portrait, Flaunt and Bounce

For a picture of one person. The face is found and the head moves as one rigid piece, so it never stretches, while the body breathes and turns in depth, swings at the hips, or bounces on the spot. The background stays still, and Erase holds any area you choose.

Depth paint

Or do it yourself. Brush size, hardness, strength and depth value over a colour-ramped overlay that actually shows what you have painted, with undo and redo.

Parallax Studio with a dusk landscape, its Auto Depth map shown as a colour overlay from blue sky to red foreground
The depth brush painting over a blue, cyan and warm colour-ramped depth overlay

Built for

Artists who want motion from still art without a 3D pipeline.

Visual novel scenes

Layered character art and backgrounds stay editable while you tune depth for subtle motion.

Short loops

Export GIF or MP4 loops for previews, social posts, trailers, or in-game presentation.

Offline by design

The depth, segmentation and pose models download once and then stay on your machine. No cloud service ever sees your pictures.

Stack

Python and Qt, with the models running locally.

UI

PySide6

Image processing

NumPy and OpenCV

AI

ONNX Runtime, Apache-2.0 models

Export

imageio and ffmpeg, optional NVENC

Download

Packages for every platform.

Fetching the latest release from GitHub...

Win Windows

64-bit. The installer is per-user, so it does not ask for administrator rights.

Tux Linux

x86_64. RPM for Fedora and RHEL, DEB for Debian and Ubuntu, or a portable tarball.

Mac macOS

Apple Silicon, macOS 12 or newer.

Src Source

Run it with Python 3.11 to 3.13 on any of the three.

The depth and pose models are not bundled. They download on first use, about 75 MB in total, and are kept on your machine afterwards.

More from Long Weekend Labs

Companion tools for visual art workflows.