What kind of theories is Uplaya.com based upon?

[KVRdataminer]

Has there been ways of quantifying the characteristics of a melody?

http://uplaya.com/

Part of the marketing pitch

"Hit Song Science™ provides immediate feedback on your song's potential for commercial success and instant legitimization in the market for high-scoring music.

Do you want help choosing the best channels and markets for your music and which of your songs have the best chance for success?

The Music Universe™ helps you understand the hit potential of your songs within different markets and niches, and includes targeted promotional features to help your music gain visibility with audiences who already like your sound.

Our rating system gives you immediate feedback on the quality of your music, its competitive edge in the music industry, and its reception among professionals and music lovers.

For example, if your song receives a Gold or Platinum Audionaut Award (Auddy™), then the underlying musical patterns are similar to songs that have been hits in the past.

The Music Universe™ includes many languages and all genres from around the world. Ultimately, The Music Universe™ will analyze and catalog all music ever created and digitized."

[Meffy]

I once wrote a story that had letter usage frequency patterns closely matching eighteen recent best-sellers. Once it gets published I KNOW it's going to make me a fortune.

Wnat theory is it based on? I think P.T. Barnum's Theory.

[sockofgold]

Wow. Just... Wow. OP, no, this is total BS.

I would never trust the scientific opinion from a website called "uplaya" to begin with, but the "theory" is total crap.

[Meffy]

To expand, 1800s showman Phineas Taylor Barnum was (wrongly) credited with having said "There's a sucker born every minute."

I'd put this right up there with poetry collection publishers who accept any entry, no matter how awful; fashion modeling and child beauty pageant scams; and those "Can you draw Peppy?" correspondence art schools who advertise on matchbook covers. It's a way to hook people on disgorging a prolonged series of expenses.

In my opinion.

[Gamma-UT]

Has there been ways of quantifying the characteristics of a melody?

http://uplaya.com/

Part of the marketing pitch

"Hit Song Science™ provides immediate feedback on your song's potential for commercial success and instant legitimization in the market for high-scoring music.

David Huron's book Sweet Anticipation goes some way into explaining theories as to what makes a good melody or harmonic progression (he stresses that he could be wrong but there's a fair amount of research to back up his position). Having said that, that's not how Uplaya works - although how it works is not really material to whether it can do what it says on the tin.

They don't go into a lot of data, but based on what they do say, Uplaya basically looks at the way frequencies are used over time in a piece of music and this is a fairly well understood branch of technology. You can do it yourself with the EchoNest API (check out the Music Machinery blog for some examples). Similarly, Mark Sadler's group at UEL in London has developed software that can analyse audio for harmonic progressions. The last time I saw/heard it was on some Beatles songs. What Uplaya does is then compare what it's found with a database of known hit songs which, apparently are subdivided by genre. I expect what happens is that it finds a lot of deep boomy kicks in RnB and so groups stuff with deep boomy kicks in that grouping.

Similarly, it's going to find that most hit tunes stick to a particular set of notes, with short excursions into nearby keys. But any competent A&R will, at least unconsciously, recognise that. But the most important part is that all this exercise tells you is what has proven to be a hit in the past - it can't tell you about the future. It can, arguably, tell you that a track is similar enough to an existing song to grouped with it. But, using statistical analysis, it can't tell you that it will be a hit as, I'd argue, that most real hits need some kind of memorable gimmick at the very least to get any form of airplay or coverage without any other kind of push. Looking at past hits doesn't do a great deal for you in that respect.

[KVRdataminer]

Thanks Gamma UT

Echonest is even more interesting. Not only analyse the music but also determine the preference and link you to other people with similar preferences.

[jancivil]

science which tells you exactly why this music is just like that other music. So you know how to market it.

WOW.

Someone knows how to bullshit no DOUBT.

[KVRdataminer]

Brian Whitman talks about The Echo Nest's "machine listening" platform and the pitfalls and promise of music data. PP presentation.

http://www.slideshare.net/bwhitman/the- ... er-21-2009

Obviously only for those who are interested in pattern recognition, ai, etc, etc.

[jancivil]

A hit song is a hit because people fell hard for it. People fall hard for something for very subtle complex inexplicable reasons.

To be a massive hit, it's managed to appeal to a mass mindset, which means at the same time it's got sameness going for it, but it has enough *personality* to stand out from the crowd. The sameness, the group mind, is like an ocean. There are waves moving in the same direction but there are many waves and many wakes from what travels on waves. It's very complex; people are complicated creatures, even when they are acting kind of the same...
The thing which stands out, is a *person*; a person is not a machine. A person has a mind, a dynamic, even a soul going for them, has weirdnesses and perversity built in. When you see, let's look at something recent, a Lady Gaga, the music is predictable. It's easy to say what the content is based on historical models. It's safe music.
There's nothing much to say about it; that is NOT why this is a hit. It's a hit because this personality meshes just right with the zeitgeist of the time in some pretty peculiar ways. That's what's the deal you're dealing in with a 'hit record'.

it's both monolithic and dynamic; it's like Borg. I do think trying to quantize this and make an algorithm to replicate it to predict it, is Utter Fail waiting to happen.

I personally find the impulse to do this kind of thing very suspect. It's a con. It's a person who readily recognizes this crass impulse in other people, to game a system like 'pop music' instead of getting their hands dirty and finding out things on the up and up.

[oneonezero]

that songs which sell good do so becaue they are similar to older ones which
also sold good, and that this means it had something to do with the musical
structure, is a complete fail of a postfaktische theorie (translation needed).

the similarity of music which sells good has probably more to do with the fact
that music which sells good is often made the same people, composed by the
same composers, published by the same publishers, or even bought by the same
customers.

but this doe snot automatically mean that it sells god because there would be
musical certain patterns which people like best.