Music Machinery

Check out Paul Lamere's talk about playlisting that he presented at ISMIR 2010 (The International Society for Music Information Retrieval has conferences about all this stuff, and Paul founded The Echo Nest, which Spotify later bought): ISMIR: The International Society for Music Information Retrieval https://ismir.net/ Finding a path through the Jukebox: The Playlist Tutorial: https://musicmachinery.com/2010/08/06/finding-a-path-through-the-jukeb… >Tutorial 4: Finding A Path Through The Jukebox -- The Playlist Tutorial. The simple playlist, in its many forms -- from the radio show, to the album, to the mixtape has long been a part of how people discover, listen to and share music. As the world of online music grows, the playlist is once again becoming a central tool to help listeners successfully experience music. Further, the playlist is increasingly a vehicle for recommendation and discovery of new or unknown music. More and more, commercial music services such as Pandora, Last.fm, iTunes and Spotify rely on the playlist to improve the listening experience. In this tutorial we look at the state of the art in playlisting. We present a brief history of the playlist, provide an overview of the different types of playlists and take an in-depth look at the state-of-the-art in automatic playlist generation including commercial and academic systems. We explore methods of evaluating playlists and ways that MIR techniques can be used to improve…

Reminds me of this old Python experiment, which will ‘swing’ any MP3: https://musicmachinery.com/2010/05/21/the-swinger/ Unfortunately most of the examples seem to be gone, but I did find this great one on YouTube: https://youtu.be/9lBP-KvujRY

I was just going through their archives trying to answer my question and found a great discussion on this post: http://musicmachinery.com/2011/05/14/how-good-is-googles-instant-mix/ There are some insightful comments from names I recognize from my research.

It's actually pretty debatable whether this year's KDD Cup will really help the science of music recommendation: http://musicmachinery.com/2011/02/22/is-the-kdd-cup-really-music-recom… Because it's entirely anonymised, not just the users but the artists too -- c.f. Netflix's problems with deanonymization: http://33bits.org/2010/03/15/open-letter-to-netflix/ This means you can't use any interesting characteristics of the music itself, or the associated metadata, to aid the recommendations. All the interesting domain knowledge is stripped out, which likely means the best solutions still won't work as well as algorithms that use metadata (like Last.fm's) or content analysis (like Pandora's) or both, and certainly won't lead to any particularly interesting insights about what drives people's tastes. Disclaimer: I work at Last.fm

Quite the opposite. The journalist idea of "Anonymous" is very different from the insider idea of Anonymous. In the simplest terms, Anonymous is anyone, really. It's not limited to 4chan users (there are other chans, and the users sometimes overlap). 4chan itself is just a convenient place to recruit people to perform an attack and later use as a scapegoat. As an example, the anti-Scientology protests were reported under the catch-all banner of Anonymous, but the people organizing it and hitting the streets were also from rival forums such as Something Awful. Funny enough, the protest movement is completely different from (and at odds with) the original catalyst: other Anons DDoSing Scientology websites for using the DMCA to take down a video on YouTube. As someone mentioned earlier in the comments, one core motivation is trolling, or "lulz," so under this so-large-its-useless umbrella term "Anonymous," there are groups sabotaging one another just because it's funny. The core people organizing* these DDoS attacks come and go, but they are a dedicated few who have the time and skills to design such programs to distribute to the masses. As in the Time magazine hack, it was respectably sophisticated (this article does it justice: http://musicmachinery.com/2009/04/15/inside-the-precision-hack). And yes, this core group does have botnets, as well as excellent social engineering skills to motivate *channers to download and execute their software. I love watching people and…

Bring back echo nest's Rosetta stone. https://musicmachinery.com/2010/02/10/introducing-project-rosetta-ston…

Your post led me down the rabbit hole of over-compression, not for the first time, and I ended up at this video which gave a very succinct answer to my question of: "Okay, I realize that music is getting compressed, but what are we losing?" https://www.youtube.com/watch?v=3Gmex_4hreQ Via a nice article with analysis: https://musicmachinery.com/2009/03/23/the-loudness-war/

Music machinery made a post[1] on what men and women listen to on Spotify. It has 40 most popular artists for either genre, 20 most gender skewed artists, and 12 most gender skewed genres. I haven't gone through and annotated each and every one of them, but at a glance it seems neither has a lot of distortion at the top 40. But in the male-skewed genres/artists there is a lot more distortion than in the female-skewed. [1] http://musicmachinery.com/2014/02/10/gender-specific-listening/

I would have to disagree with the age. 14 seems like the time when someone starts consciously listening to music. U2 was popular around the 90 to 00 and the last good album U2 created was 10 years ago. Therefore 20 and younger probably don't know them. I would like to see a pool to confirm or disprove this. Here is something interesting: http://musicmachinery.com/2014/02/13/age-specific-listening/

An engineer said they have defenses against this >Checking for offensive words is only one of the filters we have. Even without that filter, it’s essentially impossible to get recaptcha to return a false result. http://musicmachinery.com/2009/04/27/moot-wins-time-inc-loses/#comment…

More fun with sound stretching: http://musicmachinery.com/2010/05/21/the-swinger/

also, code examples: http://musicmachinery.com/2014/01/16/new-genre-apis/

The Time poll for person of the year is notorious for being rigged. In 2009, 4chan got moot to the top, in addition spelling out "marblecake also the game" with the first initials of the 21 top nominees. [1] In 2012, something similar was done [2]. 2013? Yup. http://www.dailydot.com/news/time-person-of-the-year-miley-cyrus-rigge… Anyway, my point is that the Time poll isn't even a reflection of popularity. The 2013 article also covers these 2 topics. 1. http://musicmachinery.com/2009/04/27/moot-wins-time-inc-loses/ 2. http://www.dailydot.com/society/hackers-time-person-year-kjugaschamber…

neat stuff - bit of explanation/behind the scenes is good too: http://musicmachinery.com/2013/01/02/boil-the-frog-2/

I think that what we're seeing is actually just one service being repackaged as many. Echonest provides the data supporting all of these services, so it makes sense that they all look very similar and that there is no remarkable difference in quality from one to the next. However, even with this saturation there actually are a lot of successful music discovery services, so not really a money black hole. http://musicmachinery.com/2012/12/02/music_discovery_is_not_a_lie/#com…

This is really awesome. They cut the entire song into one-beat segments of sound and video, and then categorized each beat by similarity to other beats in the song, and then after each beat there's a random chance that, instead of playing the next beat, the playback will transition to a different but similar-sounding beat from elsewhere in the song. More info at the author's blogpost: http://musicmachinery.com/2012/10/28/infinite-gangnam-style/

You should checkout PiCloud [1] or MrJob [2]. Both of these seek to make MapReduce dead simple using Python. More importantly, you can do all your testing on a Desktop PC. Then when you need real horsepower, you just tell it to "spin up on EC2." Disclaimer: I have yet to use either, but I've heard good things. [1] http://www.picloud.com/ [2] http://musicmachinery.com/2011/09/04/how-to-process-a-million-songs-in…

That reliance on "certain number of identical entries" was used by 4Chan's gaming of the Time 100 poll in 2009. The so-called "penis attack": http://musicmachinery.com/2009/04/27/moot-wins-time-inc-loses/

Dave - I like hackathons. In particular I like the 24 hour constraint. Having to build something in 24 hours (really probably 18 hours), forces me to make decisions, to get to the heart of the problem. Similarly, building something that I know I will only have 2 minutes to demo makes me really focus on what is important. I think the constraints lead to more creativity, just like the Haiku with 17 verses and 3 lines opens the door for whole libraries of verse. Some more of my thoughts on hackathons here: http://musicmachinery.com/2012/02/20/hackathons-are-not-nonsense/

Some related discussion on the Loudness Wars: http://news.ycombinator.com/item?id=1030026 http://musicmachinery.com/2009/03/23/the-loudness-war/ http://createdigitalmusic.com/2011/07/from-the-trenches-of-the-loudnes…

FAQ: http://static.echonest.com/BohemianRhapsichord/faq.html Blog post: http://musicmachinery.com/2011/11/06/bohemian-rhapsichord-a-music-hack…

It's all done algorithmically, here's a good blog post about the data: http://musicmachinery.com/2009/12/09/a-rising-star-or/

Some users from 4chan managed to hack recaptcha and embarrass Time magazine seriously at the same time not long ago. Basically they stuffed the ballots on the Time online voting page where users could vote for most influential person of the year. Most influential person ended up being Moot, founder of 4chan. Just to rub it in they managed to spell "marblecake also the game" out of the first letters of the top 21 entries. Note: This hack shows more about Time's incompetence than recaptcha's. In-depths story here: http://musicmachinery.com/2009/04/27/moot-wins-time-inc-loses/

Isn't this victim of precision hack? http://musicmachinery.com/2009/04/15/inside-the-precision-hack/

Yes... I think it's a statement about Time's approach to IT: Willing to use data from online surveys it develops, but not to keep results up to date. Maybe not funding/mandating IT to use sufficient resources to prevent result-spamming. And consider the online-journalistic issues about "Timeliness" (sorry) of Times news posts: the error's been known for hours, but Time's article (http://www.time.com/time/specials/packages/0,28757,1883644,00.html) is still starting out MARBLECAKE as of 12:30pm, MDT, despite contention at http://musicmachinery.com/2009/04/17/the-time-empire-strikes-back/ that Time was "Striking back"... and no nearby link to news about Time's getting hacked. Will they wait for Newsweek to break the story?

"...the notorious online bulletin board where hackers like to hang out." lol But seriously - here's the fun part: http://musicmachinery.com/2009/04/15/inside-the-precision-hack/

Indeed. The author is referring to the up-a-half-step modulation that Barry Manilow tactically used all over the place. I call it a "Manilow". --- But yeah, it's just one of the many elements of musical craft that's missing in today's popular music. Sampled sounds are (by definition) lifeless. Sequenced rhythm tracks are unnaturally consistent. See Paul Lamere's "In search of the click track": https://musicmachinery.com/2009/03/02/in-search-of-the-click-track/ A lack of rhythmic variation. Consider the folk tradition of wandering from the time signature; Pete Seeger and early Paul Simon for instance. And the Beatles, certainly Generally a lack of pickup notes, or "anacrusis". And the lack of an inspirational melody or chord change. --- 60's pop music was, I think, a special case because the hit songs had to get noticed and appreciated in a worst-case listening environment; typically a crappy AM radio.

Reminds me of a cool hackathon project I saw a few years ago, The Infinite Jukebox. It finds pairs of beats in a song that sound alike and during playback has a chance at each beat with of jumping to its paired match, creating a seamlessly looped version of the song with added novelty from verses and choruses having varying order and length. The original hack, done on song that works particularly well, is still up: http://infinitegangnamstyle.playlistmachinery.com The playback code often gets stuck looping the first beat when loaded for the first time or when the tab loses focus, which is fixed by reloading. It works quite well on songs with repetitive elements and instrumentation divided cleanly between beats. The original site for submitting arbitrary songs is broken, but a fork is hosted by someone else. Here with an instrumental piece it works well on: https://eternalbox.dev/jukebox_go.html?id=1LaCW0R8Q7oIY3tKtDgZOw An interesting series of posts on the creator's blog explains it works, how it interacts with the structure of pop songs, and how the parameters can be tuned for individual tracks.[1] And it's open source,[2] though the beat splitting is based on an API from (a company since purchased by) Spotify. It seems that once the initial processing is complete, the playback is purely client-side; this is probably why snapshots of the site at archive.org are fully functional, quite remarkably for such interactive content.[3] There's also a variant where one…

I am quite interested in this topic. There is some research on that field. Some links I collected in the past: http://thesis.flyingpudding.com/ http://dynampd.ubitux.fr/ http://erikbern.com/2015/09/22/presentations-about-spotify-music-recom… http://erikbern.com/2015/09/24/nearest-neighbor-methods-vector-models-… http://www.wired.com/underwire/2013/08/qq_netflix-algorithm/ http://www.slideshare.net/erikbern/collaborative-filtering-at-spotify-… http://benanne.github.io/2014/08/05/spotify-cnns.html https://news.ycombinator.com/item?id=8137264 http://www.playdar.org/ https://github.com/RJ/playdar-core http://news.ycombinator.com/item?id=3876724 https://github.com/bmcfee/librosa http://forever.fm/ http://blog.petersobot.com/introducing-forever-fm http://musicmachinery.com/2011/05/14/how-good-is-googles-instant-mix/ http://musicmachinery.com/2012/11/12/the-infinite-jukebox/ Btw., as I am developing my own music player, my main important feature/goal is a kind of automatic DJ which automatically selects music and I also especially had a discover mode in mind. However, that main feature did not evolve that much so far because there were so much other things to implement first. For now, it only supports access to files in your file system, and it looks on tags and artist and adds quite some randomness to it. https://github.com/albertz/music-player/blob/master/WhatIsAMusicPlayer…

The closest thing to a documented case I've seen is a report of people gaming the Time Person of the Year poll so that the top person was moot, and the first letters of the candidates spelt out "Marblecake. Also the game.": By understanding how reCAPTCHA worked – the team was able to double their productivity (since they usually only had to enter one word instead of two). To further optimize their voting they created a poll front-end that allowed you to enter votes quickly while giving you an update of the poll status (and since it is a 4chan kind of crowd, they also provided the option to stream some porn just to keep you company while you are subverting one of the largest media companies in the world. https://musicmachinery.com/2009/04/27/moot-wins-time-inc-loses/ However, this is slightly different as people were deliberately solving CAPTCHAs (and watching porn) rather than wanting to watch porn and also incidentally solving CAPTCHAs that they had no direct interest in.