Capturing with a PC Webcam
Requires v0.25.0-exp.1 or later.
This feature is experimental. Compared with an iPhone setup, accuracy is lower — especially for facial expression and head orientation. In the app's home screen, iPhone Camera is the recommended default.
Overview
This is a setup that captures motion using only a webcam connected to your Windows PC. You can drive the head, body, hands, and expression without preparing an iPhone.
Capture is built into the Fusion app, so there is no need to launch the Capture app separately. The Fusion app analyzes the webcam feed and sends the motion straight to the Studio app.
In an iPhone setup, ARKit measures face pose and expression directly, but a webcam setup has no ARKit. This is why accuracy is lower than an iPhone setup.
What You Need
- Windows PC
- A webcam (one whose diagonal field of view you know — you need it for the settings)
- A stand to mount the camera (one that mounts on top of your display is recommended)
No iPhone, phone holder, or WiFi setup is required. Everything runs on the PC.
How to Use
- Launch the Studio app and the Fusion app.
- In Select Capture Camera on the Remote app's home page, choose PC Webcam. The setup steps from there on switch to the webcam configuration.
- On the Capture page, turn on the Capture switch. Turning it on reveals the camera settings and everything below them.
- In the Camera section, choose the camera device and mode (resolution / FPS) you want to use.
- Set Camera FOV to the diagonal field of view of the webcam you are actually using.
- Check the video preview to confirm your face and upper body are in frame.
- Run the camera calibration on the avatar page.
Camera Settings
The items in the Camera section.
| Item | Default | Description |
|---|---|---|
| Camera Preview | — | The webcam feed. Check that your face and upper body are in frame. |
| Webcam Device | Auto | The webcam to use. If left unselected, the first device that starts successfully is used. |
| Webcam Mode | 640x480@30 | Resolution and FPS. Only the modes the selected device actually supports are listed. |
| Camera FOV | 78° | The camera's diagonal field of view (20–120°). |
| Mirror | On | Flips left and right. Matches how the iPhone (ARKit) looks. |
| Flip Vertical | Off | For correcting devices that output a flipped image. |
Always match Camera FOV to your actual device. The scale of the head's forward/backward motion is calculated from this value, so if it is off, head distance will not be reflected correctly.
Webcam spec sheets almost always list the "diagonal" field of view (e.g. Logitech C920 is 78°). The default value follows this convention.
The default is 640x480@30 because, on a USB 2.0 webcam, uncompressed 720p drops to around 10fps due to bandwidth limits. MediaPipe downscales the video further inside its model, so raising the resolution has little effect on tracking quality — securing the frame rate is the better trade.
Tracking Settings
The items in the Tracking section.
| Item | Default | Description |
|---|---|---|
| Current Provider | MediaPipeHolistic | The tracking provider actually in use. |
| Pose Model | Full | The model for body tracking (Lite / Full / Heavy). Heavier models are more accurate but slower. Shown only for MediaPipeFacePoseHands. |
| Head Depth Motion | On | Estimates and reflects the distance between your head and the camera from the apparent spacing of your outer eye corners. |
| Head Distance | 0.6m | The fixed distance from the camera to your head (0.3–2.0m). Used when Head Depth Motion is off, and as a fallback while your face is not detected. |
Providers
Two are available on Windows (since there is no ARKit, the iPhone-oriented MediaPipeAndARKit does not appear as an option).
- MediaPipeHolistic — Analyzes face, body, and hands together in a single Holistic Landmarker task. This is the Windows default chosen by
Auto. - MediaPipeFacePoseHands — Runs the three Face / Pose / Hand tasks separately. With this one, you can choose the Pose Model from Lite / Full / Heavy.
Switch using the Switch provider & restart button. The provider is only applied at app startup, so changing it restarts the app.
Checking FPS
The FPS section shows the frame rates for App / Head / Body / Hands / Webcam. If motion looks choppy, you can check here which stage is falling behind. If the webcam's FPS is low, review the mode setting (check whether the resolution is too high).
Differences from an iPhone Setup
| iPhone (recommended) | PC webcam | |
|---|---|---|
| Expression | Measured by ARKit | ARKit-equivalent blendshapes assembled from face landmarks |
| Head orientation | ARKit's transformation matrix | Estimated from 4 points: outer eye corners, forehead, and chin |
| Head distance | Measured | Estimated from the spacing of the outer eye corners, or a fixed value |
| Capture app | Launched separately on the iPhone | Built into the Fusion app |
| Connection | Over WiFi | Runs entirely on the PC |
Current Limitations
- This is an experimental feature. Compared with an iPhone setup, accuracy is lower — especially for facial expression and head orientation.
- The distance between your head and the camera is an estimate, so if the Camera FOV setting does not match your actual device, the scale will be wrong.
- The tips on camera placement and shooting environment in Getting Better Results apply as-is from the iPhone-oriented content (except for the heat countermeasures).