Can AI-driven A&R transform the music business?

Last year, YouTube’s Global Head of Music Lyor Cohen called out the record industry’s obsession with data.

He suggested that “right now everybody’s so drunk off the data that they’re putting these kids in places prior to having any artist development”.

The gut instinct vs data-led A&R debate has been raging in the record business for a while now, and some major players are heavily investing in machine learning to future-proof their A&R capabilities.

One of the early pioneers in this field was the sadly deceased Universal Music Group exec, Jay Frank, founder of Futurehit Inc., which predicted the chart potential of emerging songs.

Elsewhere, in 2018 Warner Music Group acquired Toronto-born startup Sodatone whose tech combines streaming, social and touring data with machine learning to identify the unsigned talent that it believes has the best chance of future success.

That same month, UK-based online talent scouting platform Instrumental raised over $4m to build on its proprietary digital A&R tools.

Instrumental founder Conrad Withey argued at the time that machine learning is improving, rather than hurting the tradition of signing and developing artists, writing that “it’s time to welcome the machine into the creative process – not to replace your gut instinct, but, actually, to empower it”.

If there was any doubt about Withey’s advocacy for machine learning then, the two competitive UK Christmas No.1s Instrumental scored last year and in 2018, via its artist services division frtyfve, has surely silenced any critics.

Serving as further proof of the impending impact that advancements in data analysis and machine learning will have on the industry, this year, we learnt that retail giant Amazon is patenting its own hit prediction technology. 

According to its US patent filing, the software can analyze listening data to identify ‘early adopters’ of ‘obscure’ media content, and may then subsequently be able to predict the future popularity of that content, whether it’s music, videos or even books based on the early engagement.

However – in spite of large investments in the industry, artificial Intelligence in music is still underdeveloped, but it does absolutely have the potential to transform A&R.  At least, that’s what HITLAB – a company quietly making in-roads in this field – is suggesting.

The Montreal-headquartered digital media and artificial intelligence company claims to be ‘revolutionizing the way entertainment content is discovered, produced and consumed’.

The company’s Music Digital Nuance Analysis (DNA) tool is a patented technology that the company claims is able to predict the potential success of new and unknown songs and recommend playlists to music fans.

“We want to be a major tool for AI-driven A&R.”

Michel Zgarka, HITLAB

According to HITLAB, its patented tech uses “musical pattern extraction and state of the art signal processing to isolate music attributes and transform them into quantifiable mathematical equations”.

What this means is that the software is enabled to create unique signatures for each song (which consists of 83 attributes, according to HITLAB) and then, by mining the data, hypothetically, DNA can find songs that hold similar signatures.

Furthermore, HITLAB claims that its DNA Playlist Recommendation technology is “the most sophisticated recommendation software in the world”.

This is because Hitlab’s recommendation’s system’s conclusions are based on actual sound attributes – as opposed to other systems, which use metadata.

How this works in practice, explains HITAB President Michel Zgarka (pictured), is that the company’s algorithms can identify what a Top 100 song is in any given country, or “whatever type [of song] we feed the machine”.

“It will compare any song that is presented by a songwriter or a producer, and they’ll be able to get a rating quickly of that song against whatever information we have, or data available, to compare it with,” he explains.

“It will rate that song and the producer will be able to fine tune their song to get better ratings, and they’ll know where they stand as an opportunity.

“So this could be done very quickly and it provides an opportunity for producers, or publishers or so on, to quickly identify a series of songs that could be successful in one list or another.”

HITLAB has also developed a mobile application called Can-U-Sing (CUS), a vocal scoring app that allows users to sing well-known songs and see how well their version stacks up against the original.

The app uses patent pending vocal analysis technology, and according to the promo video (see below), evaluates musical elements like the ability to hold rhythm, melodic precision and vocal technique.

The promo video also states that “recordings can be submitted to contests where singers can win cool prizes and,” – now this is where it gets really interesting – “get noticed by major record labels“.

“The innovative technology used in Can-U-Sing makes it a fantastic audition tool for singing competitions, eliminating the need for costly physical auditions,” writes HITLAB on its website.

In terms of real life implementation of its tech, one case study HITLAB lists on its website is a partnership with the Boston Red Sox, which resulted in the ‘CanUSing Your Way to Fenway Park’ contest, which offered the winner the chance to sing the national anthem at the stadium.

Finalists and the winner were chosen by collecting recordings and filtering performances using HITLAB’s app.

In April last year, HITLAB Media, a subsidiary of HITLAB Inc., signed a long-term co-development and co-production deal with Paris-based production company, WeMake, which is a division of France’s Federation Entertainment. The latter company raised over $18m in 2018 to expand into key global markets.

The partnership will see WeMake and Hitlab Media jointly produce an ‘original format’ project with the working title ‘Man vs The machine’ – The ‘machine’ being HITLAB’s patented and Patent Pending technology CanUSing and DNA tech.

Also last year, in December, HITLAB inked a long-term and exclusive licensing deal with DAKIA Digital (DAKIA), subsidiary of Dakia Global, which in turn, last year, teamed with Universal Music Group (UMG) to develop ‘a new music-focused experiential hospitality category’.

The licensing deal will see DAKIA undertake the commercialization of HITLAB’s Can-U-Sing and DNA tools within the music industry and the hospitality properties DAKIA is developing.

“The partnership with DAKIA will open international market-driven scopes to HITLAB global endeavours,” said Zgarka, following the announcement of the deal.

“We are extremely excited to be able to provide proprietary technologies that can revolutionize the way labels and artists interact with their fan base, as well as the discovery process of new talent.

“Furthermore, our technologies have the possibility to be integrated in different segments of the entertainment industry and we are looking forward to explore the TV market as well, as HITLAB develops international properties for both the theatrical and broadcast media.”

Looking to the future, Zgarka tells MBW that HITLAB wants “to be a major tool for AI driven A&R,” and, echoing Conrad Withey’s thoughts from this 2018 MBW blog, suggests that its software can help A&R professionals, rather than render them obsolete.

“We don’t want to replace the human,” insists Zgarka.

“With our AI, instead of having to listen to 1,000 songs or more a year, I can [analyse] 100,000, which will be [cut] down to maybe 100 – making sure I get the cream of the crop.”Music Business Worldwide