WITH the charts dominated by manufactured bands that all sound the same, you could be forgiven for thinking that record companies use robots as talent spotters. They might not yet, but they soon will: a Spanish company has developed software that 鈥渓istens鈥� to music and predicts how likely it is to be a hit.
Polyphonic HMI of Barcelona claims its software, called Hit Song Science (HSS), picked out the music of jazz songstress Norah Jones as destined for major chart success 鈥� months before she topped the US charts and scooped a clutch of Grammy awards.
HSS was developed to help record companies determine the hit potential of music prior to its release. And plans are afoot for a retail version of the system, which will suggest bands you might like to listen to given your current tastes in music.
Advertisement
The HSS software looks for songs that match the musical traits of known hits. Each song is run through a set of signal filters that identify and measure more than a dozen musical patterns, including melody, harmonic variation, beat, tempo, rhythm, pitch, chord progression and fullness of sound. The result is a set of raw numbers that represent each track. In a database of 3.5 million songs currently on the music market, the traits are scattered all over the map. Yet in the past five years of Billboard magazine鈥檚 Top 30 chart listings, hits were concentrated into a number of small clusters. 鈥淭here are a limited number of mathematical formulas for hit songs,鈥� Polyphonic HMI鈥檚 chief executive Mike McCready says. 鈥淲e don鈥檛 know why.鈥�
But songs with matching traits don鈥檛 always sound the same. McCready says Beethoven and U2 come out with similar values, and Norah Jones falls into the same cluster as hard-rock outfit Van Halen. Long-lived classics by artists like the Beatles and Elvis carve out their own distinct traits.
Last month, Jones and her fellow songwriters walked away with no fewer than eight Grammy awards for her first album. The accuracy of HSS has persuaded five major record labels to give it a test drive. However, having hit traits doesn鈥檛 guarantee a song will top the charts, although it does increase the odds for success. Such indicators are crucial for the music industry, which has to decide where to concentrate promotional spending.
Music shops may be able to use a variant of the software. McCready says that some retailers are keen to install the technology, which would recommend new songs and bands to customers who input details of their favourite music. The technology might be incorporated into the kiosk where you can listen to CDs before buying.
McCready hopes the HSS software will encourage labels to take risks on new artists and genres. But Peter Bentley of University College London, who specialises in making software that mimics biological processes, says it may only reinforce current formulas for success. His group has developed a program that improvises musical accompaniment by making notes swarm like insects around those of a musician. 鈥淭he music industry is not exactly renowned for its daring exploits,鈥� he says. 鈥淚f you rely on the computer too heavily, you will miss out on the new things.鈥�