A Headset in the Top Five: Engineering Notes on Dichoptic Vision Training

Valve's Steam Frame has been out a little over a month, is still heavily backordered, and has already landed as the fifth most-used headset in the Steam Hardware Survey — the first survey to include it. A standalone headset that natively supports SteamOS, off to a strong start.

Most coverage treats that as a hardware story. I'd argue it's a behavior story, and behaviors are where the interesting engineering problems live. When headsets climb the usage charts this fast, VR stops being launch-week novelty and becomes part of ordinary routines. Once a device is in the routine, the question shifts from "what can it render?" to "what can it do for me beyond games?"

For a specific group of people, one answer is vision training — and the reason it works in a headset at all comes down to hardware affordances, not content.

The constraint: two eyes, two independent displays

Amblyopia — commonly called lazy eye — is often framed as a childhood window that closes, and many adults are told there's little left to do. That framing is frustrating, and incomplete: binocular coordination and depth perception are trainable skills given the right stimuli.

The catch has always been engagement. Traditional exercises are repetitive, and patching or red-blue anaglyph glasses are visible, awkward, and easy to abandon.

What a modern VR headset gives you that a phone or TV never could: two independent eyepieces. That's a rendering primitive. You can send a different image to each eye, per frame, in sync, at headset refresh — which is exactly the substrate dichoptic training needs. No shutter glasses, no color-channel hacks.

What Amblyotube actually does with that primitive

Amblyotube, developed by Seven Sports, is recreational and educational software for visual coordination and attention exercises in VR, on Meta Quest. The core idea: instead of a dedicated exercise app, you watch YouTube-style content, and the app makes each eye's view deliberately different.

Dominant Eye Shader. The stronger eye gets a gentle, adjustable digital blur (with contrast, brightness, gamma, and opacity controls). At full opacity it acts as digital patching, but the interesting mode is partial transparency — both eyes keep receiving signal, so the brain has to reconcile them rather than defer to the dominant one.

AI Sharpener (MFBF). On the weaker eye's view, human figures (faces, bodies) are detected and marked with a bold red outline, a yellow-green highlight, and a pulsing flicker at a steady rhythm. The design intent is attention capture: a salient, moving target the lazy eye can lock onto against a busy natural video background.

Wedge Occlusion (optional). Part of each eye's view is blacked out, so each eye sees unique portions of the scene. This eliminates binocular rivalry — the two eyes competing instead of cooperating — by forcing asymmetric coverage.

Magenta Focus Cue. A bright magenta cue in the lazy eye's view, paired with a subtle neutral grey dot in the dominant eye's view. Magenta is chosen because it's rare in natural video (skin tones, greens, blues), so it survives as a distinctive signal. Tracking the merged cue is the fusion training: teaching the brain to combine both eyes into one cohesive 3D image and recover depth cues.

There's also a built-in LogMAR-style vision check — Landolt C optotypes at a virtual 3-meter distance, up to 30 trials per eye, with a clinical staircase algorithm that can auto-update the therapy profile to match detected severity. Severity presets (Mild / Moderate / Severe) tune the shader and sharpener defaults, and setup requires correctly selecting which eye is affected — pick wrong and you train the dominant eye instead.

Interface and deployment notes

A floating control panel in VR, resizable and movable, with both controller and hand-tracking input. Video comes from an integrated YouTube search bar plus an offline catalog of child-friendly content for sessions without a network. Standard playback controls, language selection, and a Reset All that restores default severity settings. Room View streams video in AR passthrough over your real room, as an alternative to the default virtual environment.

Why the hardware trend matters here

A vision-training tool only works if it gets used, and adherence is the hardest part of any repetitive regimen. That's the argument for building on content people already like: no adhesive patches, no anaglyph color distortion, no separate ritual. Put the headset on, pick a video, train for 30–40 minutes. If Steam Frame's numbers mean headsets are becoming default hardware, the ceiling on what people will use them for keeps rising.

Grounding and safety

Amblyotube is a supplementary vision training and assistive tool for binocular coordination and depth perception. It is not a diagnosis, treatment, or cure, and does not replace care from an eye care professional. Recommended use is 30–40 minutes daily, maximum 1 hour per session, ages 13 and up.

Download Amblyotube on the Meta Quest App Store: https://www.meta.com/en-gb/experiences/amblyotube/25906906972338493/