Amblyotube: Lazy Eye Training App for Meta Quest

The question regarding AI glasses replacing headphones highlights a broader truth about wearable devices: no matter how advanced the hardware is, if users do not wear them daily, nothing happens. This rule operates even more strictly in vision therapy. For individuals with amblyopia (lazy eye), what causes them to quit is rarely a lack of willpower, but rather the discomfort of adhesive patches, the dizziness of anaglyph glasses, and above all, sheer boredom. Unengaging training sessions rarely last past three weeks. From a software engineering standpoint, this is not merely a medical problem, but a user retention problem. No matter how sophisticated the algorithm is, if patients do not put on their headsets, the code is never executed.

Seven Sports approached this challenge by inverting the sequence. Instead of hunting for standalone training content, they turned the content people already watch daily into the arena for therapy. Amblyotube is a software application for Meta Quest headsets—including Quest 2, 3, Pro, and 3S—that delivers dichoptic vision and attention training while users watch YouTube-style videos. The breakthrough lies in hardware architecture: VR headsets feature independent eyepieces for each eye, allowing software to simultaneously stream entirely different video feeds to the left and right eyes respectively.

This architectural advantage opens a design space previously impossible with traditional physical patches or red-blue glasses. The system operates through distinct filtering layers. First, a Dominant Eye Shader applies adjustable digital blur, contrast, brightness, gamma, and opacity to the stronger eye, achieving up to 100% opacity for virtual patching. Second, the lazy eye view runs an AI Sharpener in MFBF mode, which identifies human figures—faces and bodies—within the video stream and selectively applies sharpness only to the weaker eye's feed. Visual accents like fluorescent yellow-green highlights or bold red outlines can also be pulsed at specific frequencies to draw the ignored eye's visual cortex back toward the target.

Stock photo for illustration only, not from the actual event

An interesting engineering tradeoff here involves focusing on human figures rather than the entire frame. While computational efficiency matters, the primary reason is the direction of human attention. Stimulating the entire screen leaves the visual system overwhelmed, whereas social targets like faces and bodies trigger evolutionary responses. Furthermore, software incorporates selective Wedge Occlusion, covering parts of both fields of view with black wedges to minimize binocular rivalry. Binocular fusion training is also supported using high-saturation magenta pink visual cues for the lazy eye and neutral gray dots for the dominant eye, prompting the brain to merge disparate signals.