RTCwebcam

RTCwebcam is a python application that can make use of a camera on other device(client) in the local area network as a virtual video device(webcam) on a linux computer(server). The application use the WebRTC API to send video stream of a client to the server, and use v4l2loopback to emulate video devices on the server.

Requirements

• openssl
  • On ubuntu-based linux distribution, it can be installed with the following command:
    sudo apt install openssl
• v4l2loopback
  • On ubuntu-based linux distribution, it can be installed with the following command:
    sudo apt install v4l2loopback-dkms
• python packages
  • list in requirements.txt
  • can install all packages with the following command:
    pip install -r requirements.txt

Running

1. First Execute run.sh

   chmod +x run.sh
   ./run.sh

2. Then open the browser on an Android, IOS devices,or other devices with cameras, and go to https://<server-ip>:8080.

   • You can reference the script output to know the possible server IP address.
   • Since a webpage can only request for camera permissions under secure connection mode, you need to use a https:// link to access the webpage.
   • However, because the ssl certificate is self-signed, you will see a warning message when you open the webpage.
   • You can ignore the warning message and continue to use the webpage.
   • A popup window will appear to ask you to allow the application to access your camera. If not, please ensure that the browser has camera permission.

3. On the webpage, you can see the following options:

   • Source: Select the camera you want to use. Useful in case you have multiple cameras.
   • Target: Select the target device on the server you want to use.
   • Streaming: Start streaming.
   • Stop: Stop streaming.

4. After start streaming, you can see the preview of the camera on the bottom of the page. On the server-side, you can now open webpages or applications that need to use the camera, like Google Meet, OBS, etc. And you will see the video device labeled RTCwebcam Device # in the devices menu.

Screenshots

• Client

• Server

Additional Information

Plugins

• The application is capable of inserting plugins to process each frame of the video stream.
• A plugin class needs to have process method that takes a frame, which is a numpy array in RGB format and have shape (height, width, 3), as input and returns a frame in the same format as output.
• Please refer to modules/Plugin.py for more information.
• Currently, there are two simple plugins avaliable
  • FlipPlugin: Flip the video stream horizontally.
  • PortraitPaddingPlugin: Padding the video stream into landscape mode. Useful for applications that will force crop the video stream in potrait mode, ex. Google Meet.
• To add and activate a plugin, you need to add to add the plugin into global_plugins list in server.py, PortraitPaddingPlugin is enabled by default.

Works with OBS virtual camera

• OBS will load v4l2loopback and create a virtual camera device at /dev/video0 (if not found) when its virtual webcam starts.
• v4l2loopback-ctl should have ability to add more virtual camera devices. However, I cannot let it work in my computer for now.
• Therefore, I chose to create more than one device while loading v4l2loopback module, and only use devices with id>=1 to avoid corrupting, so one should start the application before starting OBS virtual camera, and make sure to let the script reload v4l2loopback moudle.
• After RTCwebcam starts, one can add it as a source of an OBS Scene by selecting Video Capture Device (V4L2) in the Sources menu.

Debugging

• For now, IOS devices will stream videos in yuvj420p format, and recent version ffmpeg will produce warnings while decoding such format. As a result, the color will be somewhat different at the server-side, but still usable.
• To avoid show tones of warnings, I chose to let python log to stdout, and redirect stderr to /dev/null by default.
• If there is any problem with the application, please use ./run.sh --debug to show the error logs and open an issue.

Limitations

• The application is only works on Linux for now.
• The resolution and framerate is force to be 640x360 and 30fps for now. Though you can change the resolution and framerate in the first few lines in static/client.js, but it might not work as expected.

Test Platform

• KDE neon with ubuntu 20.04
• python 3.8

Credits

This project is modified from the following projects and links:

• aiortc/examples
• 用 Python+WebRTC 和吴彦祖视频
• Bootstrap
• jQuery

TODO

• Add more plugins and make them be able to select from the front-end.
• Make use of akvirtualcamera to emulate cameras on Windows and MacOS.
• Improve performance and add more resolution support.
• Make use of v4l2loopback-ctl to dynamically add or remove virtual camera devices.