Bottom Line Up Front (BLUF): A depth map is a black-and-white copy of a frame where brightness shows the distance to the camera: near objects are lighter, the background fades to black. This tool builds one from your video or image for free and entirely on your own device: the file is never uploaded, no tokens are charged and there is no paid API behind it. The Depth Anything V2 Small model runs right in the browser, so you see the result as it is made, and the finished video downloads as MP4 or WebM.
A depth map is the same frame redrawn in shades of grey by distance to the camera. The closer an object, the lighter the pixel; the background goes black. No textures, no skin, no faces — only the geometry and relief of the scene.
You need it in two cases. First, to get past video generation filters: models like Seedance refuse recognizable faces and film frames, while a depth map has nothing to check — there are no faces in it. Second, to transfer camera motion: a depth map shows where the camera moved and how the objects stood, and the model draws its own scene with the same movement.
The tool never sends your file to our server or anyone else's. The Depth Anything V2 Small model (Apache 2.0 licence) loads into the browser once and is then served from cache, and the computation runs on your graphics card through WebGPU. If WebGPU is unavailable, a CPU fallback kicks in — slower, but it works everywhere.
Video frames are decoded with WebCodecs, each one becomes a depth map, and then everything is assembled back into a video with the original frame rate and the highest resolution your browser can handle. The output is MP4 or WebM, whichever your system supports. No tokens are charged for this model: there is no paid request at all.
You can process both a video and a single image. Pick a file with the button or drag it straight into the window. Unusual formats such as HEIC, AVIF or TIFF are converted automatically — just like elsewhere on the site, nothing is blocked by format.
For an image the result is a PNG at the original resolution; for a video it is a clip with the same frame rate. The preview is visible while processing: frames appear as they are ready, not only after the file is assembled. A long run can be cancelled.
By default the depth map is black and white with the foreground bright. The near dark switch flips the scale — sometimes small details read better that way. For clarity there are colour maps, Inferno and Viridis: they paint distance as a gradient and help you see the planes where grey blends together.
Resolution barely affects speed: the model always computes a frame at a fixed 518×518, and resolution only matters when the finished video is assembled. The real lever is the number of frames, so a long clip on the CPU is slow: a minute of source can take several minutes.
You need a modern browser — Chrome or Edge. On a weak laptop do not start with 4K: begin with a short clip, check the result, then take a bigger file.
Choose frames with clear geometry: a street, an interior, a person in full height. Sharp changes of planes and camera movement transfer best. If you need the map for generation, make it first and then feed it to the model together with your own scene description.
Correct. The model and its weights load into the browser and the computation runs on your device. Videos and images are never sent anywhere, and there is no paid API in this model.
Nothing. No tokens are charged for a depth map: the work happens on your computer, not on our server.
Speed depends on the number of frames and on whether you have WebGPU. On the CPU a long clip takes many times longer — try a short clip or a single image.
Yes, the tool accepts images too: the output is a PNG at the original resolution.
MP4 (H.264) or WebM — whichever your browser supports. The original frame rate is preserved.