When it comes to Algolia users, we’re dedicated to providing the best quality, value and service to meet their needs. The more we discuss with them, the more amazed we are by the quality and diversity of their projects. That’s why we’re introducing today a new series of articles to highlight our favorite projects from the Algolia community.
This first project caught our eye because it tackles the hottest content trend of 2015: online video. Krish and his team at Unreel.co are working on creating the best video discovery platform on the web. We’ve invited them to tell their story on our blog in the form of a guest post.
We founded Unreel with the goal of unveiling the best video moments on the web. We came up with the unique concept of automatically crawling the internet’s videos to figure out which parts were actually worth watching. Those videos, with their worthwhile moments tagged and categorized, would then be viewable on our site, Unreel.co.
By presenting videos broken down by their ‘moments that matter,’ we had just invented a new way to discover, view, and share trending videos. Today we have collected over 3.5 million moments/scenes available for search with over 90,000 videos to watch from various channels.
Every new Beyonce video generates tons of attention and thousands of conversations. Her 7/11 video clip was no exception. What was different this time is that people were focusing more on the small details throughout the video than the song itself. Altogether, we identified more than 600 moments/scenes within the video that people most talked about, from her baby Blue Ivy making a split-second appearance, to people running in the background.
To accomplish this, we built an AI – we call it Bumblebee internally (think Transformers) – to find the most highlighted moments in digital video and bring it to the surface in a new and interesting way. Our tech looks at things like trending data, meta-data, UGC, and engagement data around videos to identify moments within viral videos that strike a cord with people.
Be it moments that make you laugh, cry, cringe, or experience any other number of emotional reactions our favorite videos can cause us. As we go out and sift through all this incredible video data, we programmatically create video channels on our platform to categorize and organize user’s video discovery experience. In upcoming versions of Unreel, we’ll let anyone create those channels to tune their viewing experience so you can watch, interact with, and share only the moments you care about.
The Secret Sauce
The key to our success is going to be our ability to empower users to customize their discovery experience. No tool is going to be more important in doing that than search. We had a classic build vs. buy challenge when it came to incorporating search into our platform. Our initial thought process was to build an in-house solution, setup a full infrastructure/framework and house our own search.
We quickly realized this would have been a huge undertaking, burning through resources we could ill afford to loose. We needed to find an alternative; an easy to integrate, customizable, instant search experience with a pay as you go model. Algolia is just that.
The Algolia Factor
Our goal is to make moments search easy – so we set the following attributes to index: moment tag, title, description and video channel. And ranked our search results based on class (moments vs videos) and rank (video score).
We discovered Algolia on Product Hunt, and have been fans ever since. The detailed documentation and tutorials made it seamless to integrate it into our platform. The whole process to go live took approx. 3 weeks – which, by the way, is super awesome by the way compared to building our own in-house solution.
We first spent a lot of time on determining how our data would look like in Algolia as a record to be optimal for search with all the data points we needed for building a unique search experience. This really helped us to make sure that we don’t waste type re-syncing/re-indexing data unnecessarily.
We then used the mongo-algolia adapter to sync all of Unreel data into Algolia. We had to make several tweaks to the adapter to make it work for us. We also built our UI in parallel to build out the interfaces of displaying search results appropriately on unreel.co
Calculating the cost of development time, resources and monthly maintenance of a search-cluster, our search integration has already paid off in a big way and is propelling our progress.
We’ve presented search to some big name partners in the video space and already received applause both for our ability to search scenes within videos but also the speed at which the videos show up. Unreel’s future is bright thanks to Algolia search.
Unreel is launching its first iOS app called JustLOL – which highlights the best scenes of the funniest videos on the internet.
Want to showcase your project? Drop us a line at firstname.lastname@example.org