Depth map

ready

Drop a video or photo here MP4, MOV, WebM · JPG, PNG, WebP · or click to choose The file is never uploaded — all processing happens on your device

0:00 / 0:00

mozgi

No file selected

Processing

Model

Output

Adjustments

Fine-tuning becomes available after processing.

Black point0.00

White point1.00

Gamma1.00

Invert

How it works

All tools

How to make a depth map from a photo or video

A depth map is a black-and-white image where brightness shows distance from the camera: lighter is closer, darker is farther. Depth maps are used for 3D parallax and "living" photos, background blur, compositing, and as input for ControlNet and other AI workflows that need a depth map. The tool computes it right in your browser with the Depth Anything V2 neural network — for photos and for entire videos, frame by frame.

How to use it

Click "Open a video or photo…" or drag a file into the window: a JPG, PNG or WebP photo, or an MP4, MOV or WebM video.
Choose a model and a mode — "Depth map" or "Contour map".
Click "Compute depth". The first time, your browser downloads the model; after that it loads from the cache.
Adjust the contrast and compare the result with the source using the "Original / Depth" toggle.
Download the result: photos are saved as PNG, videos as MP4 at the same resolution and frame rate.

Settings and tips

The Fast model (~50 MB) works for most tasks. Accurate takes about three times longer and renders edges more cleanly. Maximum quality is for complex scenes and is the heaviest.
Black and white points stretch the contrast. Move them if the map looks gray and flat.
Gamma shifts detail toward the foreground or the background.
Invert is for software that expects the reverse scheme, where white means far.
In edge mode, the points and gamma are replaced by sensitivity and line thickness.
For video, test a short clip on the fast model first, then process the whole video.

Limitations

The map is relative: it shows what's closer and what's farther, but not the distance in meters.
Glass, mirrors, water and fine details (hair, branches) often come out inaccurate.
For long videos, the map's resolution is reduced to fit in memory. The tool will let you know.
Saving video requires a browser with WebCodecs — Chrome or Edge. Audio is not carried over to the final MP4.
Speed depends on your device. With WebGPU, processing runs on the GPU; without it, on the CPU, several times slower.

Privacy

Photos and videos are not uploaded to a server: the neural network runs on your device. Only the model (from Hugging Face) and libraries (from jsDelivr) are downloaded from the internet. The site collects visit statistics: Cloudflare Web Analytics without cookies, while Yandex Metrica and Google Analytics load only if you've accepted cookies. Learn more in our privacy policy.

Other tools: turn a logo into a 3D shape with SVG to 3D, and run a small source image through photo upscaling first. For ready-made generation ideas, see the AI prompt catalog.

Depth map FAQ

Is it free?

Yes. No sign-up is needed, and there's no limit on the number of files. The only limit is your device's processing power.

Will the map work with ControlNet?

Yes. It's a standard grayscale depth map where lighter means closer, which is what ControlNet depth models expect. If your software uses the reverse scheme, turn on "Invert".

Can I make a depth map from a video?

Yes. The tool processes every frame and assembles an MP4 at the same frame rate. The video has no audio.

Why is processing taking so long?

Most often the browser is running without WebGPU, a heavy model is selected, or the video is long. Try Chrome or Edge, the fast model, and a short clip.