Recommendations should be configurable. Sometimes I'm in the mood for music that I haven't heard before. So I would want a list of music that is of a genre I don't typically listen too, but is highly rated. Also, I may want to check out music that is new to the system, or music that is highly rated, but not by many people (undiscovered).
Off the top of my head, here are some types of recommendations
1) More of what I like: Uses genre/style + ratings, plus people who rate like you
2) More of the genre I like: Uses genre/style + ratings, ignores others
3) Stuff I don't listen to but others like: Other's ratings - your ratings
4) Stuff I might like, but others haven't found: Genre + ratings, low rate count
5) New Stuff I might like: Genre + date added + ratings
We will probably want to end up producing multiple "feeds" of recommendations so that you can check out your recommendations under multiple categories.
Wednesday, January 12, 2005
Music: Artist Compensation
Artist compensation is something that is important to get right. Artists have to think they will get paid using the system, or else they wont want to use it. If I were putting my music up on the system, my main concern would be that people could circumvent the system and get my music without me getting paid for it.
Maybe we need 2 steps here. Artist need to begin to understand the disconnect between distribution and compensation. We aren't paying them based on how many copies of the song were created... If somebody emails somebody else a copy of the song, that doesn't cost me anything, and it doesn't really cost the artist anything either except lost opportunity.
So what are we really compensating the artists based on. I think song ratings is the way to go. The more favorably your song is rated and by more people, the more you get paid. I think this model makes sense as well, because the higher your song is rated, the more likely it is to be recommended to somebody, and the more value it has to the system (better recommendations for songs create more value for the user).
So the song has value to the system primarily in it's ability to both 1) bring more users into the system and 2) keep more users happy with the system (and subscribed). So the artist compensation model should be based on how successful a song (and the artist) is at doing those things.
Quantity of music doesn't matter either... if you only put one song up, but everybody loves it, it's just as valuable as somebody with 50 songs that some people like.
So practically what does this mean? Do we count positive recommendations and breakout the cash based on those recommendations?
The thing I like about that model is that it encourages the arist to throw the curve, by bringing their fans/family into the system. We should consider that when putting the payment structure together. If Bill creates four songs and puts them up in the system (at no cost to him) and gets 10 members of his family to sign up and rate his music highly, it seems like he should get some cash money out of that. It might not be much money, but it would be nice if we could put him on the map. Those positive recommendations would also potentially push his music out to others... and depending on how they rate the music, it will move up or down in the ratings (average rating).
Maybe we need 2 steps here. Artist need to begin to understand the disconnect between distribution and compensation. We aren't paying them based on how many copies of the song were created... If somebody emails somebody else a copy of the song, that doesn't cost me anything, and it doesn't really cost the artist anything either except lost opportunity.
So what are we really compensating the artists based on. I think song ratings is the way to go. The more favorably your song is rated and by more people, the more you get paid. I think this model makes sense as well, because the higher your song is rated, the more likely it is to be recommended to somebody, and the more value it has to the system (better recommendations for songs create more value for the user).
So the song has value to the system primarily in it's ability to both 1) bring more users into the system and 2) keep more users happy with the system (and subscribed). So the artist compensation model should be based on how successful a song (and the artist) is at doing those things.
Quantity of music doesn't matter either... if you only put one song up, but everybody loves it, it's just as valuable as somebody with 50 songs that some people like.
So practically what does this mean? Do we count positive recommendations and breakout the cash based on those recommendations?
The thing I like about that model is that it encourages the arist to throw the curve, by bringing their fans/family into the system. We should consider that when putting the payment structure together. If Bill creates four songs and puts them up in the system (at no cost to him) and gets 10 members of his family to sign up and rate his music highly, it seems like he should get some cash money out of that. It might not be much money, but it would be nice if we could put him on the map. Those positive recommendations would also potentially push his music out to others... and depending on how they rate the music, it will move up or down in the ratings (average rating).
Music: More on Peercasting
So I downloaded the client and am currently listening to some japanese radio station.
Unfortunately this technology will probably end up being abandoned once the record labels find out about it. They'll do their best to sue everybody out of business. Course it's an open source project so it can't really die... I'm interested in watching this one play out.
The problem with broadcasting of any kind is that you're always beholden to the copyright owners. And since most people want to listen to music owned by the record labels, they have to play by the labels rules.
Bittorrent and Peercasting are great pieces of technology for the system though, because they solve even more of the technical hurtles involved in distributing large (although they get relatively smaller every day) media files. And since we are creating a business model independent from distribution restrictions, we want to encourage this kind of development.
It's all about the artists though... enough artists have to be convinced to see the value of this kind of business model for it to take off.
Unfortunately this technology will probably end up being abandoned once the record labels find out about it. They'll do their best to sue everybody out of business. Course it's an open source project so it can't really die... I'm interested in watching this one play out.
The problem with broadcasting of any kind is that you're always beholden to the copyright owners. And since most people want to listen to music owned by the record labels, they have to play by the labels rules.
Bittorrent and Peercasting are great pieces of technology for the system though, because they solve even more of the technical hurtles involved in distributing large (although they get relatively smaller every day) media files. And since we are creating a business model independent from distribution restrictions, we want to encourage this kind of development.
It's all about the artists though... enough artists have to be convinced to see the value of this kind of business model for it to take off.
Music: PeerCast P2P Radio
PeerCast P2P Radio
Ok, this peercasting sounds really cool. I wonder if we can integate this into the system somehow. I'm thinking that we're going to want some kind of thick client component to handle stuff like managed downloads, pushing recommendations, integration into common media players, etc... Perhaps peercasting could be part of the equation.
I'm going to have to get a couple of people together to try this thing out.
I bet the crew at mixaddicts would be good test subjects.
Ok, this peercasting sounds really cool. I wonder if we can integate this into the system somehow. I'm thinking that we're going to want some kind of thick client component to handle stuff like managed downloads, pushing recommendations, integration into common media players, etc... Perhaps peercasting could be part of the equation.
I'm going to have to get a couple of people together to try this thing out.
I bet the crew at mixaddicts would be good test subjects.
simultaneous releases
Does the film industry need simultaneous releases?: "While that results in fewer CD sales, the costs to the record companies to sell music online are negligible compared to making, packaging, and shipping CDs around the world. "
Yeah, cause they're not doing anything once the music has been produced.
The article is talking about releasing DVDs and online at the same time, but I this quote stuck out to me.
Another interesting quote from the article.
This begs the question of how much such activities actually hurt the movie industry. The actual economic impact is arguably negligible — studies have shown that those who download movies are no less likely to go to the theater or rent and buy DVDs.
I gots to read them studies.
Yeah, cause they're not doing anything once the music has been produced.
The article is talking about releasing DVDs and online at the same time, but I this quote stuck out to me.
Another interesting quote from the article.
This begs the question of how much such activities actually hurt the movie industry. The actual economic impact is arguably negligible — studies have shown that those who download movies are no less likely to go to the theater or rent and buy DVDs.
I gots to read them studies.
Tuesday, January 11, 2005
Music: UI Prototype
Tonight I began working on some simple UI pages to illustrate some of the feature I've been talking about. They're not ready to be shown yet, and they wont be very pretty when I do show them off, but they will help those who are more visual learners to understand what kind of user interface to expect from the system.
It's taking a little while because I'm coding the XHTML up by hand so that it's symatically correct and can be marked up using CSS. It will make skinning and changing the prototype UI much easier, which is good because I don't expect the first version to look very good.
It's taking a little while because I'm coding the XHTML up by hand so that it's symatically correct and can be marked up using CSS. It will make skinning and changing the prototype UI much easier, which is good because I don't expect the first version to look very good.
Monday, January 10, 2005
Hyperlinks are blue and underlined because they are hyperlinks
I just removed the styling for hyperlinks from the CSS because hyperlinks are blue and underlined because they are hyperlinks.
I don't buy into this whole, "links should be almost indistinguishable from regular content" thing. Blue underlines fit into this minimalist color scheme and so by gosh, you're gonna be able to see those links.
The only think I might change are the hyperlinks in the right side menu. Those lists do a pretty good job of letting you know visually that they can be selected, but for links I insert in my posts, I want them to stand out as you're reading. They are there to encourage you to clicky clicky.
I don't buy into this whole, "links should be almost indistinguishable from regular content" thing. Blue underlines fit into this minimalist color scheme and so by gosh, you're gonna be able to see those links.
The only think I might change are the hyperlinks in the right side menu. Those lists do a pretty good job of letting you know visually that they can be selected, but for links I insert in my posts, I want them to stand out as you're reading. They are there to encourage you to clicky clicky.
Music: Song Meta-Data or Tracks vs. Albums
So I've been thinking about song meta-data (I think dictionary.com has popups, but I wouldn't know because I use firefox).
Meta-data is going to be important to the song searching/recommendation process. In order to know whether we should recommend a song to someone, we'll need to know enough about it to classify it in relationship to other songs.
I spent some time looking at the list of fields in theCDDB
It seems like their database is designed for albums primarily, while I'm inclined to make tracks the primary piece of data in the system. Albums can definitely be defined on the system, however they are used as a device of the artist to build a collection of songs. Since the system isn't tied to the typical album model, it seems like we should take advantage of that. And along with throwing out the baby, we can throw out some bathwater... for example: Artists will no longer be encouraged to produce "filler" tracks to bring their "album" up to 50 minutes. If you have 3 really good songs, put them up on the system. Group them up as an EP if you like, but those three popular songs could bring you as much income as 4 full albums worth of material that isn't as appealing.
But at the same time, we need to make sure to support artists that already have an "albums" worth of material put together that they want to classify as a "unit".
What do y'all think. In an idea world, with no legacy system to support, how should your music be organized?
Meta-data is going to be important to the song searching/recommendation process. In order to know whether we should recommend a song to someone, we'll need to know enough about it to classify it in relationship to other songs.
I spent some time looking at the list of fields in theCDDB
It seems like their database is designed for albums primarily, while I'm inclined to make tracks the primary piece of data in the system. Albums can definitely be defined on the system, however they are used as a device of the artist to build a collection of songs. Since the system isn't tied to the typical album model, it seems like we should take advantage of that. And along with throwing out the baby, we can throw out some bathwater... for example: Artists will no longer be encouraged to produce "filler" tracks to bring their "album" up to 50 minutes. If you have 3 really good songs, put them up on the system. Group them up as an EP if you like, but those three popular songs could bring you as much income as 4 full albums worth of material that isn't as appealing.
But at the same time, we need to make sure to support artists that already have an "albums" worth of material put together that they want to classify as a "unit".
What do y'all think. In an idea world, with no legacy system to support, how should your music be organized?
Music: PodCasting defined
I'm starting a new protocol. I'm going to preface any post titles relating to the music system with "Music:" Since I can't setup categories in blogger, this will be the next best thing. If you are only interested in the system, you can ignore anything not prefaced with Music:
Here is a wikipedia link to podcasting for those that have heard me use the term but aren't familiar with it.
Podcasting is the concept I think we could use for music delivery in the system. The system would have a set of heuristics it would use to determine what songs a user of the system would find interesting. And then setup a custom podcast for that user that would deliver music to that users inbox. Podcasting is a good fit from a technology perspective because it gives distribution control to both the user and the producer. Since we populate the RSS feed, we can effectively "push" new music to the client, but at the same time, users could simply disable the client if they didn't want to get new music.
Here is a wikipedia link to podcasting for those that have heard me use the term but aren't familiar with it.
Podcasting is the concept I think we could use for music delivery in the system. The system would have a set of heuristics it would use to determine what songs a user of the system would find interesting. And then setup a custom podcast for that user that would deliver music to that users inbox. Podcasting is a good fit from a technology perspective because it gives distribution control to both the user and the producer. Since we populate the RSS feed, we can effectively "push" new music to the client, but at the same time, users could simply disable the client if they didn't want to get new music.
CSS Tweaks
I tweaked on the CSS just a little bit to get a slightly different style. The minimalist template was a little too minimal for me.
Style modifications
I'm changing to a simpler style. This one seems to have better organized CSS. Should be easier to make modifications.
You should read this blog using an RSS aggregator anyways. ;-)
You should read this blog using an RSS aggregator anyways. ;-)
Bloggers Atom Feed
If you're following the system you can safely ignore this post, unless you're interested in CSS.
Check out the atom feed for this blog. First notice that it's an XML feed. If you view source on the document, you can see the atom xml format. Second, notice that it's a pretty darn good looking XML feed. Here's another example of an xml feed. See how this one isn't as nice looking.
I haven't figured out what part of the XML/HTML specs allow you to markup XML using CSS the way these blogger guys have, but it sure is nice. If they included a link to the comments in the atom XML feed I would almost rather use that as my blog style.
Check out the atom feed for this blog. First notice that it's an XML feed. If you view source on the document, you can see the atom xml format. Second, notice that it's a pretty darn good looking XML feed. Here's another example of an xml feed. See how this one isn't as nice looking.
I haven't figured out what part of the XML/HTML specs allow you to markup XML using CSS the way these blogger guys have, but it sure is nice. If they included a link to the comments in the atom XML feed I would almost rather use that as my blog style.
Producers and Studios
I've identified another area that music labels help out artists. They help connect producers with artists, as well as having access to networks of recording studios and mastering houses.
Recording is being pushed "down" in terms of the initial investment required to start a studio, but the skills are still a pre-requisite.
The system will probably need a way for producers to enter in and offer up their services to those in the system.
I can think of 2 ways to arrange that.
1) Artists use the system to search for producers that could help them with a project. They arrange to pay for the producers service outside of the system.
Or the more interesting one
2) Producers become associated with a song in the same way that an artist is associated with a song. That way, a producer can be paid through the system, using a similar system (with different percentages probably) to how artists are compensated.
Recording studios are a different beast though, because they are typically paid by the hour, regardless of how popular the song becomes. I don't see the system having a huge impact on recording studios, except that it would give them a good networking tool. We will need to figure out whether they should pay for access to the system, or inversely, somehow encourage them to take on projects originating from the system.
That's an interesting viral technique... when somebody from the system records a song in a recording studio, that studio gets listed through the artist.
Just an idea... I think having access to those resources will be a service of the system, and as the system hosts we'll probably have to put some time into finding recording studios.
Recording is being pushed "down" in terms of the initial investment required to start a studio, but the skills are still a pre-requisite.
The system will probably need a way for producers to enter in and offer up their services to those in the system.
I can think of 2 ways to arrange that.
1) Artists use the system to search for producers that could help them with a project. They arrange to pay for the producers service outside of the system.
Or the more interesting one
2) Producers become associated with a song in the same way that an artist is associated with a song. That way, a producer can be paid through the system, using a similar system (with different percentages probably) to how artists are compensated.
Recording studios are a different beast though, because they are typically paid by the hour, regardless of how popular the song becomes. I don't see the system having a huge impact on recording studios, except that it would give them a good networking tool. We will need to figure out whether they should pay for access to the system, or inversely, somehow encourage them to take on projects originating from the system.
That's an interesting viral technique... when somebody from the system records a song in a recording studio, that studio gets listed through the artist.
Just an idea... I think having access to those resources will be a service of the system, and as the system hosts we'll probably have to put some time into finding recording studios.
Media player integration
I'm stuck here at work for a few extra minutes waiting for Gwen so I thought I'd blog a little.
One idea I had about the system was that to get real value, we need to integrate with the popular media players. iTunes seems like a no-brainer in terms of support. Integrating with iTunes would give us tight integration with the iPod as well. Similar to how iPodder or the other podcasting tools integrate with iTunes and hence the iPod.
What are some other candidates.
Is music match jukebox still a popular one?
I'm thinking that integrate into Windows Media Center Edition would be cool. The market is probably kind of small for it today, but there's potential there.
I wonder what Winamp's latest version looks like?
I've mostly been using iTunes on the mac lately because of the iPod.
One idea I had about the system was that to get real value, we need to integrate with the popular media players. iTunes seems like a no-brainer in terms of support. Integrating with iTunes would give us tight integration with the iPod as well. Similar to how iPodder or the other podcasting tools integrate with iTunes and hence the iPod.
What are some other candidates.
Is music match jukebox still a popular one?
I'm thinking that integrate into Windows Media Center Edition would be cool. The market is probably kind of small for it today, but there's potential there.
I wonder what Winamp's latest version looks like?
I've mostly been using iTunes on the mac lately because of the iPod.
Thursday, January 06, 2005
Tired tonight
I worked late tonight so I'm kind of fried. Brain storming is now a drizzle... a brain drizzle...
Not enough bits in this byte: User experience
Not enough bits in this byte: User experience"I want them to be smart enough to know (pretty much) what I want to listen to when I switch them on."
Fred has done some brainstorming on his ideal user experience. It's definitely an interesting read.
I have no idea how to build that kind of solution. But I have a feeling it could be built on top of the system. There are so many details to put something like that together.
Let's call that kind of experience "the system 10.0" :) My favorite part is the telepathic interface. ;)
Fred has done some brainstorming on his ideal user experience. It's definitely an interesting read.
I have no idea how to build that kind of solution. But I have a feeling it could be built on top of the system. There are so many details to put something like that together.
Let's call that kind of experience "the system 10.0" :) My favorite part is the telepathic interface. ;)
Wednesday, January 05, 2005
Artist community
Here's another quick thought.
Part of the value add of the labels is connecting artists and producers so that the producers can help the artists produce music. Similarly, they help connect recording studios and mastering facilities.
I think the system could be used to connect these people together as well. In addition to providing a direct feedback loop between the subscribers and the artists.
We would need to do some leg work to come up with studios and mastering houses to recommend, but that's just part of the whole relationship building we'll need to do.
Part of the value add of the labels is connecting artists and producers so that the producers can help the artists produce music. Similarly, they help connect recording studios and mastering facilities.
I think the system could be used to connect these people together as well. In addition to providing a direct feedback loop between the subscribers and the artists.
We would need to do some leg work to come up with studios and mastering houses to recommend, but that's just part of the whole relationship building we'll need to do.
More Genres
Jon posted in a comment here talking about how one system does more advanced genre work for music.
I really like that idea. I'm thinking of how we could get the benefits of something like that.
Maybe it's not a genre so much as a description of the music. And a piece of music should have multiple descriptions associated with it. Some examples of descriptions could be: contemplative, intense, melancholy. While a piece of music might only have one genre or subgenre, another level of description would be helpful in classifying the music.
Meta-data is going to be one of the more critical pieces of the system, so doing some brain storming on how to extend the "standard" musical meta-data is good.
I really like that idea. I'm thinking of how we could get the benefits of something like that.
Maybe it's not a genre so much as a description of the music. And a piece of music should have multiple descriptions associated with it. Some examples of descriptions could be: contemplative, intense, melancholy. While a piece of music might only have one genre or subgenre, another level of description would be helpful in classifying the music.
Meta-data is going to be one of the more critical pieces of the system, so doing some brain storming on how to extend the "standard" musical meta-data is good.
User experience
Let's brainstorm user experience some...
The community interaction aspect of the system points towards a web based interface. But the iTunes like seamless interaction would probably require a thick client for the desktop.
Perhaps a combination of the two... a thick client that manages your downloaded music and schedules new downloads....
ooohhhhhh.... maybe the desktop client could be a podcast tool. oh oh oh, I like that idea. Part of the service could be to subscribe to "podcasts" of recommended music. Every night your desktop app automatically downloads the latest songs that are recommended for you.
With all the podcasting hubub, I'm actually a little concerned about making that idea public... but oh well.
Ok, so let's describe the user experience for that.
You browse around the site, and download some songs you think you'll like (based on comments and reviews). You like some of the songs, and add them to your personal genre. After enough time you've built up a big enough list of rated music that the system can begin guessing at your tastes.
At this point you install the desktop app that watches the service for songs to downloaded based on recommendations.
If Adam Curry has taught us anything though, it's that podcasts use a ton of bandwidth. Perhaps the thick client can use an implementation of bittorrent to distribute the files.
This actually ties into keeping subscribers around. Imagine if after rating enough netflix movies, extra movies just ended up in your mailbox because netflix thought you would like it. And if you don't, then you can just send it back, at no extra cost to you.
The community interaction aspect of the system points towards a web based interface. But the iTunes like seamless interaction would probably require a thick client for the desktop.
Perhaps a combination of the two... a thick client that manages your downloaded music and schedules new downloads....
ooohhhhhh.... maybe the desktop client could be a podcast tool. oh oh oh, I like that idea. Part of the service could be to subscribe to "podcasts" of recommended music. Every night your desktop app automatically downloads the latest songs that are recommended for you.
With all the podcasting hubub, I'm actually a little concerned about making that idea public... but oh well.
Ok, so let's describe the user experience for that.
You browse around the site, and download some songs you think you'll like (based on comments and reviews). You like some of the songs, and add them to your personal genre. After enough time you've built up a big enough list of rated music that the system can begin guessing at your tastes.
At this point you install the desktop app that watches the service for songs to downloaded based on recommendations.
If Adam Curry has taught us anything though, it's that podcasts use a ton of bandwidth. Perhaps the thick client can use an implementation of bittorrent to distribute the files.
This actually ties into keeping subscribers around. Imagine if after rating enough netflix movies, extra movies just ended up in your mailbox because netflix thought you would like it. And if you don't, then you can just send it back, at no extra cost to you.
Subscribe to:
Posts (Atom)
