blurcam

Virtual webcam with AI background blur for Linux

One command to blur your background on Discord, Zoom, Google Meet, or any video app.

Works on Asahi Linux (Apple Silicon), Raspberry Pi, Ubuntu, Fedora, Arch, and more.

Install

curl -sSL https://raw.githubusercontent.com/jolehuit/linux-blurcam/main/scripts/install.sh | bash

That's it. The installer:

• Installs uv (fast Python manager) if needed
• Installs the v4l2loopback kernel module for your distro
• Creates the virtual camera at /dev/video10
• Starts the background daemon that auto-detects when apps use the camera

Usage

In your video app (Discord, Zoom, Meet...), select "BlurCam" as your camera.

That's it. Blur starts automatically when an app uses BlurCam, and stops when the app closes.

Your real webcam only turns on when needed - no manual commands required.

Adjust blur strength

# See current settings
blurcam config

# Change blur (applies immediately, no restart needed)
blurcam config --blur 45

Blur values: 15 = subtle, 35 = default, 55 = strong, 75 = very blurry

How to test

1. Open your video app (Discord, Zoom, Meet...)
2. Select "BlurCam" as your camera
3. You should see yourself with blurred background

Known Limitations

Switching cameras in Electron apps (Discord, Vesktop, etc.)

Due to how V4L2 works on Linux (only one process can use the webcam at a time), switching from BlurCam to your real camera may show a black screen.

Workaround: Close the camera settings menu, wait 1-2 seconds, then reopen it and select your camera.

This happens because Electron apps don't release the previous camera fast enough when switching. It's a timing issue between the app and the daemon.

Uninstall

blurcam uninstall

This removes config files, cached model, and systemd service.

Manual install

If you prefer to install manually:

1. Install v4l2loopback

Fedora / Asahi Linux:

sudo dnf install akmod-v4l2loopback
# Asahi only: sudo dnf install kernel-16k-devel && sudo akmods --force

Ubuntu / Debian / Raspberry Pi OS:

sudo apt install v4l2loopback-dkms v4l2loopback-utils

Arch Linux:

sudo pacman -S v4l2loopback-dkms

2. Load the kernel module

sudo modprobe v4l2loopback devices=1 video_nr=10 card_label="BlurCam" exclusive_caps=1

3. Install blurcam

# With uv (recommended)
uv tool install blurcam

# Or with pipx
pipx install blurcam

# Or with pip
pip install --user blurcam

4. Run

blurcam

Auto-start (default)

The installer automatically enables the daemon at boot. The daemon is lightweight - it only watches for app connections and starts the webcam when needed.

Manual control

# Check status
systemctl --user status blurcam

# Stop daemon
systemctl --user stop blurcam

# Disable auto-start
systemctl --user disable blurcam

v4l2loopback config (for manual install)

echo "v4l2loopback" | sudo tee /etc/modules-load.d/v4l2loopback.conf
echo 'options v4l2loopback video_nr=10 card_label="BlurCam" exclusive_caps=1' | \
    sudo tee /etc/modprobe.d/v4l2loopback.conf

Troubleshooting

"Virtual camera not found"

The kernel module isn't loaded:

sudo modprobe v4l2loopback devices=1 video_nr=10 card_label="BlurCam" exclusive_caps=1

"Could not open webcam"

Another app is using the camera. Close other video apps first.

Check system status

blurcam-setup

Low FPS

On slower devices, reduce blur strength:

blurcam config --blur 15

Advanced options

# Detection sensitivity (if edges are cut off or background leaks through)
blurcam config --threshold 0.4   # More inclusive (keeps more of you)
blurcam config --threshold 0.6   # More strict (cleaner edges)

# Use different webcam
blurcam --input 1

# Use different virtual camera
blurcam --output /dev/video20

How it works

1. Daemon watches the virtual camera device using inotify
2. When an app opens BlurCam, the daemon starts processing
3. ONNX Runtime runs the AI segmentation model on CPU (ARM64 compatible)
4. OpenCV captures your webcam and applies background blur
5. pyvirtualcam + v4l2loopback sends the result to the virtual camera
6. When the app closes, the daemon stops processing (webcam turns off)

The key: using the ONNX model directly instead of MediaPipe Python bindings (which don't support ARM64).

Compatibility

Platform	Status
Asahi Linux (M1/M2/M3)	Tested
Raspberry Pi 5	Not tested yet
Raspberry Pi 4	Not tested yet
Ubuntu ARM64	Not tested yet
Fedora ARM64	Not tested yet
Arch ARM	Not tested yet
x86_64 Linux	Not tested yet

License

MIT License - see LICENSE

Credits

• ONNX Community - Exported segmentation model
• Astral - uv package manager
• v4l2loopback - Virtual camera kernel module
• pyvirtualcam - Python virtual camera library