Digital Live Business Media The Great Escape 2018

CMU Great Escape Special: Meet the innovators

By | Published on Wednesday 16 May 2018

Abstract polygonal background

The CMU Great Escape Special was a magazine distributed at The Great Escape 2018 featuring articles linked to the three conferences CMU Insights presented there and other CMU reporting and research. Here is one of the articles linked to The AI Conference.

The AI Conference will feature seven innovative companies who are using data, algorithms and machine learning to change the way music is monitored, marketed and made. To get things started, we asked each innovator to describe both their product and their tech in their own words.


is using audio-ID technology to make radio monitoring faster, cheaper and more comprehensive.

THE PRODUCT: WARM is an online platform and app where everyone can monitor their songs as they are played on radio stations around the world in realtime. As a user, you upload an MP3 file of your track onto and – once processed – we will automatically track all your radio airplays of this song.

Where other radio monitoring platforms focus on large stations, WARM aims to track all music radio stations in the world. Currently we are tracking over 25,000! Once we are tracking your song on all those stations, all your airplay data is provided in realtime, including duration, country, city and station name.

WARM is a service built primarily for the long-tail of the music industry – artists, bands, managers, independent labels and booking agencies. I wanted WARM to be as affordable as possible, with a price that means everyone can use it.

THE TECH: We create a unique audio fingerprint of the MP3 file that is uploaded and this runs in a loop on all the stations all the time. Whenever there is a match, the song data and information about the airplay will be visible on the user’s dashboard. This shows within 15-20 secs after the airplay. The core technology is based on a special patented algorithm.



is busy moving the fan conversation over to Facebook Messenger.

THE PRODUCT: POP makes it easy to build and engage an audience of fans or customers using Facebook Messenger.

For many people Facebook Messenger (or ‘Messenger’) is “just another messaging app” – similar to WhatsApp, iMessage or even plain old text messages. In fact Messenger is a powerful tool to engage fans or customers at scale, offering a compelling way to communicate using interactive content and story-telling. At POP we have built a simple platform with an easy-to-use content management system that lets you do cool stuff on Messenger.

“Traditional” social media like Facebook, Twitter and Instagram promised a lot. Suddenly you could talk directly to your audience, create exciting content and get it right in front of the people who might be interested in your event. But over time, this promise has been eroded. Congested social feeds have meant platforms like Facebook, Twitter and even Instagram have brought in algorithms to sort content and determine who sees what. It’s difficult to reach fans.

Messenger is different because once fans connect with you on Messenger, they actually get the messages you send – no filter getting in the way of your content – and we typically see sky-high open rates on messages – 80-90% or higher.

This is all pretty new stuff though – in fact, Facebook only opened up the Messenger API in April 2016. This means a lot of people are still getting to grips with what you can do using Messenger, and what works well.

A lot of people think that the power of Messenger is about building “chatbots” and AI-powered experiences. But the reality is that as soon as you ask users to type things to a chatbot, the failure rate is huge. We’ve focused on the other side of what messaging is about – which is delivering a great way to tell stories, and reaching fans at scale with narrative flows and interactive messaging.

THE TECH: We build directly on top the Facebook Messenger API, and are an official Facebook Messenger Platform Development Provider, so are fully compliant with Facebook’s Messenger policies.

In terms of actually building your messaging flows and interactions, we’ve focused on simplicity as much as possible. To begin to create a flow of messages, you simply click, add some text or media, and save. These flows of messages can be used in two places – as the first interaction someone sees when they connect with you on Messenger, but also as a way to deliver onward content.

You can deliver interactive content through Messenger and direct people down different routes using ‘quick reply’ buttons. Quick reply buttons are one of the key features of automating Messenger – they function either like menu buttons or prompts for content. Quick replies are powerful because we find that the less typed input required from a user, the better the experience.

We have a component picker that lets you add rich media – images, gifs, video. Native video works very nicely in Messenger and offers a great user experience for fans. Or you can create simple content cards – a great way to link to a YouTube video, Spotify/Apple Music track or back to a piece of content on a website.

Once you’ve created your messaging flow you can review privately with your own Messenger account to ensure everything is working the way you want it, and then push it live – either as the first interaction for people on your Messenger channel, or as live update to fans who have already connected.



has been busy building tools that utilise years of historical user data to recommend shows and events, including a ‘smart gift finder’.

THE PRODUCT: Given the breadth of events we list as the UK’s biggest live guide, visitors come to us for suggestions and inspiration as well as info and tickets for specific shows. Our Event Affinity Engine fuels a range of new tools, including our Smart Gift Finder. The Gift Finder takes a few pieces of info about the user – or a friend or relative of theirs – cross-references them with what we already know about the fans of thousands of acts, and provides a tailored, relevant list of shows happening near them.

Because we work with most major ticket agents, we can then direct fans to tickets regardless of which trusted agent has them, and we’re free to switch between agents as various allocations run out, keeping us agile and reliable.

The Engine also forms the heart of a number of other new features, including one that combines data from any music fan with our wealth of data about the UK’s fan landscape, suggesting the ideal festival for each person, but also which acts on the line-up they’re most likely to enjoy. Perfect for an event like The Great Escape, where fans may be faced with hundreds of bands that are new to them and want to know who to watch.

These features mean an end to the generic, blunt gift/festival guides that currently exist – eg ‘Gifts for Dads’, ‘You like Ozzy so why not try Download’ etc. Our Event Affinity Engine connects fans to events they’ll love, as well as presenting relevant festivals to fans in the best possible light.

THE TECH: Ents24 has over a decade of archived user behaviour data consisting of millions of tracks (when a user explicitly registers their interest in a band, venue or event in order to receive email alerts), ticket transactions, competition entries, clicks, views and so on. We also have demographic data such as age and location for many of our 2.8 million registered users.

Secondly, we have the UK’s most comprehensive events database, maintained by a team of experts, complete with rich entity-relationship connections. We represent these connections natively within a graph database, and this allows for much more interesting and rapid inference during modelling when compared with flat datasets which rely solely on collaborative filtering or matrix factorisation, for example. This wide search space means that we can identify and recommend small, local events as well as the major ticketed concerts.

We combine both collaborative and content-based models to form a hybrid recommender system, the parameters of which we then tune using expert domain knowledge and user feedback – the former derived from over fifteen years of industry experience, and the latter is provided by sampling from our two million monthly users.



is providing talent scouting tools by crunching the latest data from Spotify.

THE PRODUCT: Instrumental’s music scouting platform provides daily insights into how artists and tracks are performing on Spotify. The user can filter by a series of defined parameters (eg genre, location, social reach) to identify fast emerging independent and unsigned artists.

THE TECH: The Instrumental platform ingests Spotify public API data, specifically thousands of influential playlists. Discovered tracks – and the artists creating them – are subsequently pulled into the system. Our data updates on a daily basis allowing clients to see, in near real-time, spikes in artist or track growth. Currently, we are monitoring 4.9m tracks from over 400,000 artists.



is using AI technology to create tools that can compose music.

THE PRODUCT: Our site lets users choose a bunch of high-level musical settings – such as genre, mood, length (down to the second), speed (down to the exact BPM), and where in the track a climax should occur – and create unique, rights-cleared tracks that fit those settings.

Every track you generate is original and unique, and is cleared to use globally on all platforms. Once you’ve created a track, you can go back and edit it, doing things like changing the duration – and you can download your finished tracks, which are free if you’re an individual creator and are happy to credit us where you end up using the music.

THE TECH: We use deep learning – specifically, artificial neural networks – training a system to be able to generate original melodies and chord sequences on demand. When you request a new track, our neural networks generate a totally new musical score, which is then automatically produced by our custom-built audio engine. The finished MP3 or WAV file is sent back to you, with the whole process taking between 20 seconds and a couple of minutes.



is using AI to automatically edit music to fit to video.

THE PRODUCT: MXX has created the world’s first AI engine which ‘atomises’ and re-constructs music to video instantly.

Our first product, MXX Audition Pro, is a desktop application based on a patented AI that has learned to atomise and re-construct existing music. It allows creators everywhere to use a simple, totally intuitive desktop application, to add amazing soundtracks adapted to their videos. Users import any video into the application and create a brief for how the music should accompany the video action using simple icons.

They then search connected libraries with millions of licensable tracks where any track can be dropped into the application and auditioned against the brief until the ideal track for the video is found. Once the perfect track is identified, the track self-edits through the AI engine to match the brief.

MXX Audition Pro makes it easy to synchronise events such as video edits or effects. It allows the quick auditioning of tracks against a brief. Production companies can cut or fade seamlessly between sections of music with no perceptible edit in the audio output.
It also means, for the first time ever, that anyone in the world can create awesome soundtracks for their videos.

Now, anyone can simply create a brief against short form video with intensity rises, fades, dropouts and climax points, load any track, and immediately listen to the track edited to the brief. This makes responding to the demands of today’s media fast and simple. What typically takes hours or days of a music editor’s time, now takes minutes of anyone’s time.

THE TECH: Firstly, AI is used to perform music information retrieval on audio files. The AI detects the beat and bar mapping, creating time signatures, before segmenting the work and working out where there are valid entry and exit points for each section. The sections are then meta-tagged depending on their overall intensity and the AI also tags the intensity development.

Secondly, artificial intelligence genetic algorithm techniques evolve a solution given a piece of music and a brief. The brief and music are both balanced alongside each other, even if the music does not contain components requested in the brief, or if the brief asks for impossible components; either way, the AI algorithm will find an intelligent solution. In the case of user generated content and song editing, AI techniques edit the piece using constraint satisfaction mechanisms.



is using AI to automatically create video to accompany music.

THE PRODUCT: Rotor is a purpose-built tool that simplifies the video creation process for the music industry. Using proprietary technology, Rotor enables musicians, labels and services to create promotional video content in minutes. Users add their music, choose some clips and choose a video style. Rotor will then generate a full preview of their video so they can decide if they like the results or not.

THE TECH: There is a lot of proprietary technology at work behind the scenes: We have built an engine that can extract features from each individual song and make it’s own editing decisions based on a particular video style. This is similar to how a video producer or editor would work.


READ MORE ABOUT: | | | | | |