Pixlane is loading locally. No upload, no signup, and your files stay on your device.
Upload an image or video, or use live camera mode, to estimate relative scene depth and render a colorized depth map directly on each frame.
All processing happens entirely on your device. Your files never leave your device — no upload, no server, no signup required.
Recovering depth from a single image is inherently ambiguous — a small nearby object and a large distant one can project to exactly the same pixels. The tool resolves this the way modern computer vision does: a deep neural network trained on large datasets of images paired with ground-truth depth learns the cues people read unconsciously — perspective, occlusion, relative object size, texture gradients, and shading. The network runs a single forward pass over your frame and outputs a per-pixel depth map, which is then normalized and colorized so near and far surfaces are easy to read at a glance. Because one photo carries no absolute scale, the result is relative depth — what is closer versus farther — not a measurement in meters. The same estimator is applied frame by frame for video and live camera input, so motion is handled as a sequence of independent per-frame estimates rather than a single 3D reconstruction.
Use a still image when you want the cleanest, highest-detail result; depth follows edges and surfaces best on sharp, well-lit photos, so a low-light boost pass first helps if your source is dim or noisy. Video mode runs the same estimator across every frame for a moving depth visualization, and live camera mode does it in real time for quick experimentation. In the colorized output, the gradient encodes ordering rather than exact distance — read it as a heat map where one end of the scale is the closest geometry and the other is the background. Because the model infers structure from learned priors, the hardest cases are reflective surfaces, transparent glass, and flat textureless walls, which may show softer or less certain depth.
A depth map is a useful building block for many creative and analytical tasks. Common uses include:
Everything runs entirely in your browser on your own device — your images, videos, and camera frames are never uploaded to a server. That privacy guarantee matters most for the footage people actually have on hand: home interiors, faces, and personal moments you would not want sent to a cloud API. It is also free, needs no signup, leaves no watermark, and starts instantly with no model account or quota to manage.
It predicts relative depth from an image or frame and renders a colorized map so you can quickly see which regions are nearer or farther away in the scene.
Yes. Depth Estimation now supports uploaded videos and live camera mode, so you can preview the colorized depth map over time directly in the browser.
Yes, Depth Estimation on Pixlane is completely free. No account, no signup, and no watermark on downloaded files.
Object detection and segmentation identify what is in the scene. Depth estimation instead predicts scene geometry and relative distance, helping you understand spatial layout from a single image.