Video Recommendations for Mux Video and Player, powered by Mux Data

August 22, 2022 (10 months ago)

Build your own “Up Next”: recommendations in Mux Player powered by Flike and Mux Data

By John and TobiasPartnerships 5 minute read

Mux is proud to partner with Flike, the AI-Powered Recommender API that leverages Mux Video to transcode and stream videos, Mux Player to provide an engaging “lean back” viewing experience, and Mux Data’s streaming exports of video views to inform the recommendations algorithm to surface the most engaging content for each viewer. As content creators, studios, and licensors continue to compete for eyeballs, connecting viewers to the content they find engaging is paramount. Flike learns and adapts to user behavior to help you bring the personalization experience that has previously only been accessible to companies like YouTube, Spotify, and Netflix.

Link42IconHow Flike builds recommendations 🧪

  • Video recommendations are computed using a variety of carefully selected and tuned machine learning algorithms (e.g., collaborative filtering and deep learning).
  • The recommendations are updated in real time; starting from the first click to view and continuing through any following interactions, the system retrains itself while generating new recommendations.
  • Flike’s solution is highly scalable and capable of serving thousands of recommendations per second.

Link42IconHow it works 🪧

Flike leverages video engagement information streamed instantaneously from Mux Data to classify user actions and intents as likes or dislikes. These data are used to train collaborative filtering-based algorithms, which will then recommend uniquely engaging content for each user in real time. Collaborative filtering classifies items that a user might like based on the behavior of similar users. It’s a simple yet effective way to recommend videos users will like by analyzing their engagement history, finding other users with similar engagement, and discovering their overlapping and nonoverlapping patterns. This system is built on the principle that people who evaluated items similarly in the past are likely to do so again in the future.

To put it simply, collaborative filtering does not require any information about the items or the users themselves; instead, it recommends items based on users’ past behavior. Most recommender systems of large content companies, such as Netflix, YouTube, and Spotify, are based on collaborative filtering techniques.

Link42IconHow to integrate 🧩

With the Flike + Mux partnership, we connect directly to your tech stack, allowing you to embed state-of-the-art personalization into your website or app in no time; you don’t need to spend months on data engineering, model training, and deployment difficulties inherent to most recommendations offerings. The only steps left to get the recommendations running at your site are:

Step 1: Receive credentials to the Flike Recommender API

Step 2: Create a real-time streaming data export to Flike from the Mux Dashboard

Step 3: Install the Flike plugin in your Mux Player instance

Step 4: There is no step 4 — that’s it!

Link42IconStep 1: Getting credentials to the Flike Recommender API

Flike will provide you with a customer ID and API credentials to connect your data and get recommendations tailored to your use case. Fill out this form (it’ll only take a minute), and your API key will land in your inbox on the same day.

Link42IconStep 2: Create a real-time streaming data export to Flike from the Mux Dashboard

Flike learns about video viewing behavior with a combination of in-player events and by analyzing engagement data received by Mux Data’s streaming exports. To get these data to Flike, you will set up a video view streaming export to Flike's Pub/Sub system, which starts the process of exporting all user interactions on your website to Flike in real time.

  1. Select the Environment you want to send view data from.
  2. Select Google Cloud Pub/Sub.
  3. In the Pub/Sub topic name field, enter “projects/goflike/topics/mux-ingest-${customerId}”, replacing `${customerId}` with your Flike customer ID provided in Step 1.

Link42IconStep 3: Install the Flike plugin to Mux Player

The Flike plugin hooks into Mux Player to automatically set up personalization behind the scenes. Once attached, the plugin pulls the viewer’s user ID from the player and auto-plays the next recommended video (if you want to enable that feature).

All you need to get started is to install the Flike plugin using npm i –save @flike/mux-player-plugin and add a few lines of code to the page where you want to display your Mux Player instance:

import FlikeMuxPlayerPlugin from '@flike/mux-player-plugin'

const muxPlayer = document.getElementById('my-mux-player')
const recommender = FlikeMuxPlayerPlugin.attach(muxPlayer, {
  apiKey: 'public.ABCDEFGHIJKLMNOPQRSTUVWXYZ', // provided in Step 1
  autoplayNext: true,
  autoplayWaitDuration: 3,

// fetch 4 recommeded videos
const nextVideos = await recommender.recommend(4)
// [ { assetId: 'xyz', probability: 0.4804032 }, ... ]

Link42IconStep 4: There is no step 4 — that's it!

Mux customers interested in learning more about using streaming data exports to drive real-time recommendations in Mux Player can check out Flike's demo here. The demo showcases how the new Mux Player interacts with real-time streaming exports from Mux Data to build data-driven user journeys in real time.

Create an account

No credit card to start. $20 in free credits when you're ready.

Get started for free

Subscribe to the Mux newsletter

Read more like this