The companion carries the 2D to 3D filter now

Turning a flat film into stereo 3D used to be something you set up. The converter was a separate program you compiled yourself, with a machine learning runtime and a depth model to fetch beside it, and the instructions for doing that on three operating systems were longer than anybody should have to read to watch a film.

It ships inside koko Companion now. Install the app and the filter, its accelerator runtime, the depth model and the upscaling models are already there. Nothing to download, nothing to point the app at.

What it does

Every frame is passed through a depth network, which estimates how far away each pixel is, and the picture is then re-projected into a left eye and a right eye. You are not watching a flat rectangle floating in a room any more: the scene has depth, and your headset renders it the way it renders a film that was shot in 3D.

It runs on your graphics card, and how fast depends entirely on which one you have. The companion measures your machine and tells you whether it can do this while you watch, or whether the film should be converted ahead of time instead.

Converting ahead of time

If your PC is not fast enough to convert during playback, it does not have to be. Point the Convert tab at the films you want, leave the machine on, and they are waiting in 3D when you put the headset on. A converted file plays like any other 3D film after that, with no GPU cost at all.

Where it runs

Apple Silicon uses the Neural Engine, Windows and Linux use the GPU through DirectML or WebGPU. There is no NVIDIA-only build and nothing to install for your particular card.