opentok-camera-filters

Library which lets you add visual filters to your OpenTok Publisher.

Demo

Blog Post

Browser Support

• Chrome 51+
• Firefox 49+
• Safari 11+

These filters require the Canvas captureStream API which works in Chrome 51+, Firefox 43+ and Safari 11+ (and Safari on iOS 11). Adding audio to the stream only started working in Firefox 49+.

Usage

You can view the source code for the demo for an example of how to use this library.

const filters = require('opentok-camera-filters/src/filters.js');
const filterFn = require('opentok-camera-filters');

Then you get your media strea you want to filter and pass it to the filter function eg.

const publish = OT.getUserMedia().then((mediaStream) => {
  // Initialise with filter none
  filter = filterFn(mediaStream, filters.none);

  const publisherOptions = {
    // Pass in the canvas stream video track as our custom videoSource
    videoSource: filter.canvas.captureStream(30).getVideoTracks()[0],
    // Pass in the audio track from our underlying mediaStream as the audioSource
    audioSource: mediaStream.getAudioTracks()[0],
  };

  const publisher = OT.initPublisher('publisher', publisherOptions, handleError);
  filter.setPublisher(publisher);

  return publisher;
});

Then when we have successfully connected we publish the publisher to the Session.

session.connect(TOKEN, err => {
  if (err) handleError(err);
  publish.then(publisher => {
    session.publish(publisher, handleError);
  });
});

If you want to change the filter you can use the change method, eg.

filter.change(filters.red);

Available Filters

A lot of the filters were taken from tracking.js.

red

Give the video a red tint

green

Give the video a green tint

blue

Give the video a blue tint

invert

Inverts the colour in every pixel of the video.

grayscale

Converts a colour from a colorspace based on an RGB color model to a grayscale representation of its luminance.

sepia

Applies a sepia tone to the image.

blur

A Gaussian blur (also known as Gaussian smoothing) is the result of blurring an image by a Gaussian function.

sketch

Computes the vertical and horizontal gradients of the image and combines the computed images to find edges in the image.

face

Does face detection using clmtrackr and draws an image on top of the face.

Custom Filters

If you want to create your own custom filter you just need to create a function that looks like one of the functions in the filters.js file. These functions accept a videoElement and a canvas parameter and they take the data out of the videoElement which is rendering the unfiltered video from the camera and they draw it onto the canvas after applying a filter. It should return an object with a stop method which when called will stop the filter from processing. For example creating a simple filter which draws a new random colour every second would look something like:

const randomColour = () => {
  return Math.round(Math.random() * 255);
};

filter.change((videoElement, canvas) => {
  const interval = setInterval(() => {
    const ctx = canvas.getContext('2d');
    ctx.clearRect(0, 0, canvas.width, canvas.height);
    ctx.fillStyle = `rgb(${randomColour()}, ${randomColour()}, ${randomColour()})`;
    ctx.fillRect(0, 0, canvas.width, canvas.height);
  }, 1000);
  return {
    stop: () => {
      clearInterval(interval);
    }
  };
});

You can also use the filterTask which handles transforming image data from the videoElement and just lets you pass it a filter function which takes ImageData and transforms it returning new ImageData. The invert function is a good example of a simple filter which uses this.

If you want access to the face tracking data from clmtrackr you can use the face() filter and pass in your own renderer function like so:

filter.change((videoElement, canvas) => {
  return filters.face(videoElement, canvas, positions => {
    // Do something with the positions and draw something on the canvas
  });
});

The positions are the response from clmtrackr.getCurrentPosition(). The glasses filter is an example of a face filter.