.png)
1 month ago
7
Most 3D media collapses into a regular video once it lands on a phone or laptop. Interfaces often ask people to click and drag, but in practice most viewers are not looking to steer the camera. They want to be carried by the story. This is the same comfort gap that held back 360 video on phones. It can shine in venue scale installations like the Sphere in Las Vegas, yet on personal devices it rarely fits the lean back habit of watching.
Headsets avoid this problem. Presence comes for free because a tiny shift in your chair changes what you see. The challenge is that most viewing still happens on flat screens.
Use head tracking to estimate eye positions, then reproject the scene so the camera view follows the viewer. The result is motion parallax that makes the content feel anchored in your space without asking the viewer to drag or scrub.
As you move, the content readjusts, which feels like the scene is in the room with you. There is no stereoscopy, yet the brain still gets strong depth cues from parallax, perspective, and occlusion. On any device with a front facing camera this creates an intuitive, hands free way to experience 3D.
In informal tests with nearly one hundred people, the interaction was instantly understood. No instructions were needed.
People default to lean back viewing. They accept gentle, automatic interactivity that respects the narrative, and they reserve active control for rare moments. Head coupled perspective honors that pattern. It adds presence without demanding effort, and it keeps editorial control with the creator.
Detect facial landmarks and iris centers from the front camera.
Estimate a six degree of freedom head pose relative to the screen
Compute a view matrix that treats the viewer as the virtual camera origin.
Reproject the 3D scene each frame, with temporal smoothing to manage micro jitter.
We call this viewing paradigm Window Mode because the display functions as a fixed window into the scene. Your eyes define the viewpoint on that window. When your head shifts, we move the virtual camera to the matching position and render through the same window. The math just makes that intuition exact. We estimate head pose relative to the screen, set the camera origin to your eye position, and reproject so parallax and occlusion behave the way they would if you were looking through glass. This is why it feels natural without instructions.
This approach builds on the idea of head coupled perspective popularized 17 years ago by Johnny Lee with the Wii Remote. Modern on device models and GPUs now make it practical on everyday hardware.
Low latency is critical. The shorter the delay between moving your head and seeing the updated image on screen, the more solid the world feels. If that delay is too long, the scene lags and looks wobbly. Filtering out jitter and rejecting outliers keeps edges steady instead of swimming. Respect privacy by processing face data on device and discarding it per frame. Provide a quick toggle and a graceful fallback to standard playback.
Try the live demo: https://lab.true3d.com/targets
It will ask for camera access.
We also converted a few familiar clips into 3D to show how the effect changes a shot. Start with Steamboat Willie and turn on Window Mode.
If you want to create your own 3D clips, use the free tool here.
There you can drop any mp4 and get a 3D clip out. We also have an API.
The stream and player run on True3D’s APIs. Our pipeline uses volumetric video with voxels and gaussian splats to deliver view dependent rendering efficiently. For the web we provide a drop in player component and simple APIs so you can bring this effect to your own app, game capture, or live render from Unity or Blender.
Join the Discord and say hello. We’ll help you get set up with our APIs and would love to see what you’re building.
