Foveated rendering is a graphics technique that renders full detail only in the small region where the eye is looking and progressively lower detail in the periphery, matching how human vision works. In XR displays & optics within extended reality, it is one of the main ways to drive high-resolution headset displays without a proportional rise in graphics processing cost.
The technique exists in two main forms. Fixed foveated rendering assumes the user looks at the center of the lens and reduces detail toward the edges; eye-tracked foveated rendering, also called dynamic foveated rendering, uses eye tracking to move the high-detail region wherever the gaze goes.
Why foveated rendering matches human vision
Foveated rendering works because visual acuity is concentrated in the fovea, a small pit in the retina densely packed with cone cells. Sharp vision covers only about the central one to two degrees of the visual field, and acuity falls steeply with distance from that point: a few degrees away it is already a fraction of the central value, and in far peripheral vision it is very low, although sensitivity to motion and flicker remains.
A headset that renders its whole field of view, often around 100 degrees or more horizontally, at full resolution spends most of its pixel work on regions the eye cannot resolve. Foveated rendering redistributes that work: the foveal region gets full shading, a surrounding parafoveal region gets moderate quality, and the periphery gets coarse shading. Done correctly, the user does not notice the difference.
The table shows a typical division of the image into quality zones.
| Zone | Approximate angular extent | Typical shading rate |
|---|---|---|
| Foveal | Central few degrees around gaze | Full resolution |
| Parafoveal | Ring out to roughly 10 to 20 degrees | Half resolution in each axis |
| Peripheral | Remainder of the field of view | Quarter resolution or lower |
Fixed and eye-tracked foveated rendering
The two forms differ in where the high-detail region is placed and how aggressive the reduction can be. Fixed foveation is cheap and works on any headset, but because the user’s gaze often moves away from the lens center, the savings must stay modest. Eye-tracked foveation can shrink the high-detail region much further, because it always follows the gaze.
| Property | Fixed foveated rendering | Eye-tracked foveated rendering |
|---|---|---|
| Hardware needed | None beyond the GPU | Eye tracking cameras |
| High-detail region | Fixed at lens center | Follows the gaze |
| Typical savings | Moderate | Substantially larger |
| Visible artifacts | Blurry edges if users look sideways | Rare when latency is low |
| Main risk | Over-aggressive edge reduction | Tracking latency or failure |
The fixed approach also benefits from lens optics: headset lenses already blur and distort the outer image, so lower detail there is often hidden. Savings for both methods vary widely with content and settings; published measurements range from modest gains to more than halving the shading work, so claims should always be judged against a specific application.
Rendering methods that implement foveation
Foveation is implemented by reducing how often pixels are shaded rather than by lowering geometry detail. Several GPU techniques can do this, and engines choose among them according to hardware support.
| Technique | How it reduces work | Notes |
|---|---|---|
| Variable rate shading | Shades blocks of pixels once | Widely supported on desktop GPUs |
| Fragment density maps | Varies render density across the image | Common on mobile XR chips |
| Multi-resolution rendering | Renders zones into separate targets | Older approach, more complex |
| Lens-matched shading | Matches density to lens distortion | Exploits optics, not gaze |
| Foveated streaming | Encodes the gaze region at higher bitrate | Used in cloud and wireless streaming |
In practice, developers rarely build foveation from scratch. Platform SDKs, the OpenXR extensions for foveation and engines such as Unity and Unreal expose levels such as low, medium and high, and some runtimes apply fixed foveation automatically.
Latency and other requirements for eye-tracked foveation
Eye-tracked foveation only works if the high-detail region arrives before the eye can notice its absence. During a saccade, a rapid eye jump, vision is briefly suppressed, which gives the system a short window to move the foveal region. If the total delay from eye movement to updated image is too long, users perceive blurry patches that trail behind the gaze.
| Requirement | Why it matters | Symptom when missed |
|---|---|---|
| Low gaze-to-photon latency | Region must follow saccades | Blur noticed after eye jumps |
| Accurate calibration | Region must sit on the true gaze point | Sharp area slightly off target |
| Robust tracking | Glasses or reflections can disturb it | Fallback to fixed foveation |
| Smooth zone transitions | Hard edges attract attention | Visible rings or shimmering |
Systems therefore predict gaze landing points, enlarge the foveal region when confidence drops and fall back to fixed foveation when tracking is lost.
Limitations and common misconceptions
A common misconception is that foveated rendering reduces display resolution. It does not: the panel still shows all its pixels, and only the number of distinct shading calculations changes. Another is that savings scale directly with the reduced area; in reality, geometry processing, post-processing and CPU work are largely unaffected.
- Periphery is motion-sensitive: coarse shading can cause shimmer or flicker that the eye notices.
- Text and fine lines at the edges degrade first, so interfaces should stay near the center.
- Benefit depends on the bottleneck: CPU-bound applications gain little.
- Eye tracking raises privacy questions, because gaze data reveals attention.
Understanding it simply
Your eyes see sharply only in a tiny spot where you are looking. Everything around it is blurry, even though the brain fills in the gaps so well that you rarely notice.
Foveated rendering uses that fact. The headset draws a crisp picture only where you look and a rougher one around it. Because the rough part falls where your eyes cannot see detail anyway, the image still looks sharp to you, while the computer does much less work.
Applied in practice
A studio developing a detailed flight simulator for a standalone headset struggles to hold a steady frame rate over dense city scenery. The cockpit instruments must remain sharp, but the mobile GPU cannot shade every pixel of both high-resolution displays at full quality.
The team enables eye-tracked foveated rendering through the platform SDK, keeping a full-quality region around the gaze and reducing shading to a quarter in the outer periphery. They also widen the foveal zone slightly whenever tracking confidence drops.
The application reaches its target frame rate with headroom to spare, testers report no visible blur when glancing between instruments, and the studio uses the saved performance to raise texture quality in the cockpit.