Bottom Line Up Front (BLUF): The tool turns an ordinary video into stereoscopic 3D (SBS, Half-SBS, anaglyph) and adds a virtual camera move based on the depth map. Frame depth is computed by Depth Anything V2 Small right in your browser, depth maps are smoothed over time, and the stereo views are assembled by a WebGL shader. The file is never uploaded, no tokens are spent, and there is no paid API involved.
The tool takes an ordinary video and, from the depth map of each frame, builds a 2.5D scene: near objects shift more than far ones, and from that it produces two views — one for the left eye and one for the right. That is how the video becomes stereoscopic 3D.
You can also move the camera virtually: left and right, forward and back (dolly/zoom) and slightly change the viewing angle. This is not a re-shoot — it is a re-projection of the already captured frame using depth.
The source video is never uploaded: frame depth is computed by Depth Anything V2 Small through ONNX Runtime Web (WebGPU, with a CPU fallback), and the stereo views are assembled by a WebGL shader right in your browser. There is no paid API here and no tokens are spent.
To keep depth from flickering from frame to frame, depth maps are smoothed over time, so the stereo parallax stays even and does not shimmer.
SBS 3D — two full-size frames side by side (left eye on the left, right eye on the right): good for 3D TVs and most VR players.
Half-SBS — each eye is squeezed horizontally to half width and placed side by side: the format VR players and headsets expect.
Anaglyph — red-cyan: the red channel comes from the left eye, green and blue from the right. You watch this result with ordinary red-cyan glasses, no special screen needed.
Camera movement is limited automatically: this is depth-based 2.5D, not a real 3D scene, so beyond a certain shift stretched edges and large holes start to appear. The tool does not let you go past that limit.
Small holes that appear when the camera moves (revealed background that was not visible before) are filled with neighbouring pixels, so the picture stays intact.
Nowhere. The video is decoded, depth is computed and the stereo views are assembled entirely inside your browser on your device.
No. It is depth-based 2.5D: the image is re-projected using depth, but you cannot fully walk around an object — that is why the camera range is limited.
The tool is completely free and no account tokens are spent.
Popular video formats are supported, including MP4, WebM and MOV. For fast on-device processing, clips up to a few minutes long and up to 1080p are best.