Motion-to-photon latency is the time between a head movement and the moment the display emits light showing the matching change in the virtual view. Also called tracking latency in everyday use, it is one of the most important factors in XR comfort, because a visible lag between movement and image is a direct trigger of VR motion sickness and of the feeling that the world is swimming.
Widely cited design targets put acceptable motion-to-photon latency at below about 20 milliseconds, and modern headsets use prediction and late image correction to bring the perceived latency close to zero for head rotation, even when the full rendering pipeline takes longer.
Where motion-to-photon latency comes from
Motion-to-photon latency is the sum of every delay in the pipeline from sensor to screen. The headset must sample its motion sensors, estimate a pose, let the application simulate and render a frame, compose and correct it, send it to the panel and wait for the panel to light the pixels. Each stage adds time, and the stages are only partly parallel.
The table breaks the pipeline into its typical stages. The values are approximate and vary with the device, refresh rate and application load.
| Pipeline stage | What happens | Typical delay |
|---|---|---|
| Sensor sampling | IMU measures rotation and acceleration | Under 1 millisecond |
| Pose estimation | Sensor fusion with camera tracking | About 1 to 2 milliseconds |
| Application and rendering | Game logic and GPU draw the frame | About one frame, 8 to 14 milliseconds |
| Compositing and correction | Runtime warps and combines layers | About 1 to 3 milliseconds |
| Scan-out and display response | Panel receives and lights the pixels | About 2 to 10 milliseconds |
Without compensation, these stages could easily add up to 30 milliseconds or more, which most users notice as a lag when turning their head quickly. The compensation techniques described below are what make modern headsets comfortable.
Why latency affects comfort
Latency affects comfort because the vestibular system in the inner ear reports head motion almost instantly, while the eyes see the corresponding image change only after the delay. This sensory conflict between balance and vision is a major cause of nausea, dizziness and eye strain in XR, alongside artificial locomotion and the vergence-accommodation conflict.
Head rotation is the most demanding case. When the head turns at a moderate speed of around 100 degrees per second, a delay of 20 milliseconds shifts the image by about 2 degrees, enough for objects to appear dragged along with the head. Small errors that would be invisible on a monitor become obvious because the display is fixed to the head.
Sensitivity differs strongly between people. Experienced users with VR legs tolerate more, while newcomers and people prone to travel sickness react to smaller delays. Latency problems also accumulate: a session that feels fine for ten minutes may cause discomfort after forty.
Techniques that reduce perceived latency
Headsets reduce perceived latency with prediction and late correction rather than relying only on faster rendering. The runtime predicts where the head will be when the frame is displayed and renders for that pose, then corrects the image again just before scan-out using the newest sensor data.
| Technique | How it works | Limitation |
|---|---|---|
| Pose prediction | Extrapolates head motion to display time | Errors when motion changes suddenly |
| Asynchronous timewarp | Re-projects the last frame for new rotation | Does not correct position fully |
| Positional reprojection | Uses depth to shift the image for translation | Artifacts at object edges |
| Frame synthesis | Generates intermediate frames from motion vectors | Ghosting on fast or thin objects |
| Low-persistence display | Lights pixels only briefly each frame | Requires bright, fast panels |
These techniques correct head rotation very effectively, but they cannot fully hide latency in moving hands, controllers and animated objects, which still appear slightly behind the user’s real movements when the application misses frames.
Motion-to-photon latency compared with related measures
Motion-to-photon latency is often confused with frame rate and frame time, but they measure different things. A high frame rate shortens one stage of the pipeline, yet a headset can run at 120 hertz and still show high latency if its tracking or display stages are slow.
| Measure | What it describes | Typical target in VR |
|---|---|---|
| Motion-to-photon latency | Movement to visible light | Below about 20 milliseconds |
| Frame rate | Frames rendered per second | 72 to 120 frames per second |
| Frame time | Time to render one frame | About 8 to 14 milliseconds |
| Pass-through latency | Camera capture to display in MR | Kept as low as possible, often tens of milliseconds |
Practical guidance for developers and users
For developers, the most effective step is to hold a stable frame rate at the headset’s native refresh rate, because dropped frames force the runtime into heavier reprojection. Profiling tools that show frame timing and reprojection rates should be part of every test session.
- Keep a frame budget margin: aim well below the frame time limit to absorb spikes.
- Avoid camera smoothing: never filter or delay the head pose, because added smoothing directly increases latency.
- Test on real hardware: simulators do not reproduce display and tracking delays.
- Offer comfort options: snap turning, teleport locomotion and comfort vignettes reduce sensory conflict when latency spikes.
For users, the main levers are adequate PC or headset performance, a stable wireless connection for streamed PC VR, and lower graphics settings when frame rates drop. Wireless streaming adds encoding and network delay, so a wired or dedicated wireless link generally produces lower latency than a busy home network.
Understanding it simply
Motion-to-photon latency is how long the picture takes to catch up when you move your head. Your inner ear notices the movement at once. If the picture follows a moment later, your brain gets mixed signals.
The key takeaway is that a tiny delay matters greatly inside a headset. Good headsets predict your movement, so the picture seems to move exactly with you.
Applied in practice
A studio testing a VR racing game notices that several playtesters feel unwell within ten minutes, although the game reports an average of 90 frames per second. Frame timing tools reveal regular spikes to about 25 milliseconds when other cars enter the scene, forcing the runtime into frequent reprojection.
The team also finds that the camera script applies light smoothing to the head pose to reduce jitter, adding roughly a frame of delay. Engineers remove the smoothing, move car physics updates off the render thread and reduce shadow quality for distant vehicles.
After the changes, frame times stay below 10 milliseconds with headroom, reprojection becomes rare and the latency measured with a photodiode test drops noticeably. In the next playtest, most testers complete 30-minute sessions without discomfort, and the studio adds a comfort vignette option for sensitive players.