Finnish startup Hyperlive claims its algorithm can predict which songs will become hits — with astounding accuracy.
After just seven months on the platform, Luis Fonsi and Daddy Yankee’s single, ‘Despacito,’ became the most-watched video on YouTube. Last August, the music video reached 3 billion views. Just two months later, it crossed the 4 billion mark.
Who would’ve predicted that ‘Despacito’ would’ve overcome ‘Gangnam Style’ and ‘See You Again’?
According to one music tech startup, its new technology may have.
Hyperlive has allegedly developed an algorithm that predicts a song’s hit potential — simply by using its ‘audio signature’. The technology doesn’t focus on analyzing similarity to past hits. Nor does it rely on factors like social media activity. Instead, it models a range of neurobiobehavioral responses to music as well as underpinning psychological processes.
This, says the company, allows for prediction of large-scale musical engagement with purportedly “unmatched levels of precision.”
Speaking on the results, Hyperlive CEO Geoff Luck told Digital Music News,
“The major benefit of our algorithm lies in how accurately it forecasts a song’s hit potential prior to it being exposed to a wider audience… What’s more, from artists, publishers and labels to music repositories, streaming platforms and music tech companies, we envision a whole host of value-creation possibilities enabled by an algorithm that’s able to predict a track’s potential for success.”
So, with such large claims, what proof does the company have?
Plenty, apparently. Hyperlive’s algorithm correctly predicted how 200 tracks from 10 major artists, including Ed Sheeran and Taylor Swift, would do. Since their release, the songs have accrued a combined 180+ billion streams and over 1.2 billion single sales.
Analyzing each track’s audio signature, Hyperlive’s algorithm predicted actual performance with 84% overall accuracy.
In addition, for tracks incorrectly identified as hits, predicted total track sales and streams fell within an average of 25% of their actual range.
Luck says that from a creative point of view, Hyperlive may help songwriters and producers make better hits. One key thing to note, however, is that Hyperlive has yet to show definitive proof. Its press release doesn’t contain any published graphs nor research to back up Luck’s extraordinary claims.
Until then, Hyperlive’s alleged successful new technology remains only a fairy tale. But one we’ll keep watching.
Featured image by InfoWire.dk (CC by 2.0)
Hi there, would you be interested in my fancy new snake oil?
I’ll trade you my air guitar, lightly used.
Wow, they predicted that tracks from Ed Sheeran and Taylor Swift would be hits? Where do I sign??
I wonder if their algo predicted just how hard they are going to be sued by Nike for using Hyperlive.
If this algorithm really does not rely on similarities, it is very interesting. Btw. according to their website it says they have been analysing 200 tracks from 10 major artists, not 10 tracks as it says in this article. It makes a huge difference. And please do not be so stupid bitching about analysing the major artists. If the analysis has been done from the audio fingerprint, the algorithm does not know whose song it is. 😀 So if it can predict, not just measure, this is a great tool for labels and artists.
Every other Finn is an engineer, no wonder they believe the answer is in the machine.
This would be like leaving the music choices to the one kid who would not, cannot dance.
If a publicly traded company misses a projected earning on its report by 14%, it’s a disaster. Streaming and trade charts (ie. Billboard, Spot etc.) are bogus measurement tools for artist/band popularity; that’s like consuming empty calories. Credible popularity measurement, the protein, is in live concert ticket sales, artist merchandise sales, live performance appearances and active[ation] followers.