leer

Home  |   Features   |   Private use    |   Business use   |   Download  |   Links  |   Sitemap 
 

 

 

The Moodmixer Principle ?

(released May 10. 2004 09:44:00 am)

The Moodmixer™ principle – imagine you had a machine that selected music automatically as you would select it yourself!

As soon as a digital music listener has gathered more than a few soundfiles on his PC- he likes to categorize his files – either by sorting them into folders, or if given the possibility, by assigning properties or rating them. The only purpose of this exercise is to find songs easier when trying to set up a playlist.

But when it comes to putting music together for playlists, all the work done on administering the music files doesn't help a lot because still in most cases all songs have to be selected by hand.

The newest development has thankfully gone a stage further: One can search for the assigned ratings/properties and thus collect songs with the same attributes a lot quicker.
The result is: you can automatically produce a playlist from your favorite songs, from your second best songs or from a specific genre etc. But is this really enough?

There seems to be a difference between young listeners who really like to play the same song over and over again and love to make crucial decisions on which songs to play next and the growing number of older listeners who simply like to enjoy a good music mix without having to put too much work into it.

Moodmixer is the perfect solution for an automated music selection:

• Base of playlists are songcollections either put together by hand or by using the extensive search functions but also the whole library can be used
• Playlists can be produced as non-stop play, for a certain length or scheduled
• The user can add personal weighting and rating to each song which is then used for the playlist generation- the weighting indicates the amount of times a song gets played, other ratings narrow down the selection
• With a few settings and a touch of a button Moodmixer automatically selects the music- constantly varies the mix and takes care of when and how songs are repeated.

In detail:

At PWsoft we decided to use the already known technology and combine it with a brand new selection method in order to specifically satisfy the needs of older (>25) users.

The known technology:
First of all Moodmixer lets the user decide from what kind of pool it is allowed to select songs from.
This could be the whole music library, this could be a predefined amount of songs- put together more or less by hand in Moodmixer (obviously supported by all kinds of search functions) or this could be a dynamic list derived from a search for specific criteria that can be combined ( i.e. Search for all songs from 1950-1979 with either genre "Rock” or "Pop" or you could search for all songs that contain the word "Xmas" either in filename, title, albumname or foldername and put them together with all christmas songs and songs containing "santa" in their title= christmas list). These dynamic lists would change as soon as other matching songs are added to the library. All lists are saved and can be used as basis for a playlist.
The selection method:
At PWsoft we thought that a rating system should be the basis for the amount of times a song should be played.
Music collectors tend to forget many songs they own as soon as the amount of soundfiles a user owns exceeds a certain value.
In Moodmixer the listener can rate and profile a song and this rating is taken into account when producing playlists automatically:
The higher the weighting of a song the more it gets played. Sounds easy but is quite tricky. Imagine you had 10 songs in a playlist- one weighted ten times higher than the others- now how often would you like to have this song played?

The developers of Moodmixer have put all their knowledge and experience from working for more than 20 years in the radio music business into a smart algorithm that shows a high sensitivity for the listener’s desires but also for his tolerance for repeated songs. The smart algorithm not only uses the preference weighting of each song but takes also many other factors into account. The result is: even if you like a song very much- it won’t get overplayed. A little added random factor at the right time always makes it impossible to tell the next song. Often enough the user might be surprised what kind of really nice music he owns especially if the playlist basis is the whole music library.

Moodmixer selects songs from all weighting values and thus produces highest variation in the playlists. While listening to those playlists the user will always feel that his preferred songs occur often enough but the playlists always show highest possible variation no matter how long they are or how many songs the used title pool/predefined collection or searchlist contains.

Moodmixer is based on a system of 5 criteria of which the first reflects the preference for a song; the other four can be defined individually.
All five criteria can be used to narrow down the amount of titles Moodmixer is allowed to choose from- i.e. you could have a property called ‘age’ and thus narrow down the possible music selection to music only of the 50s, 60s and 70s. Or you could let Moodmixer only play the latest stuff, you could narrow down the tempo or you could select what kind of mood the music should reflect. Or you could match the music to a situation. And of course you can narrow down the selection to your most favorite titles also. The clue is here- as soon as the mood changes you can simply adjust your slider settings, press the Mix-button and the next song will already match the new mood.

So you either have the choice of setting up a special list without the need to narrow it down or you use a wider range of music and use Moodmixer’s slider controls to exactly control the selection.

In each range of a certain property Moodmixer’s smart algorithm takes care that- if possible (meaning there are matching songs available) - value-jumps in a certain property are as low as possible. In the case of tempo this would mean that a fast title isn't followed by a very slow one. Or in the case of age a brand-new title isn’t followed by a very old one.

On top of that Moodmixer can apply an optional set of rules:

  • 1) Don't repeat a song in sequence
  • 2) Don’t repeat an artist in sequence
  • 3) Don’t play a song if it has been played around the same time the previous day (a set prior day delta indicates the amount of time a songs is being blocked)
  • 4) Reduce pool to all songs with a length longer than xyz secs and shorter than xyz secs
  • 5) If the users have added BPM-values to their songs: Songs in sequence shouldn’t differ more than xyz in BPMs (here the user can add a step value which is added if Moodmixer doesn’t find any matching songs.)
  • 6) As soon as Moodmixer finds that some of the rules cannot be obeyed because of a lack of matching songs- it loosens up the rules until it finds songs to select again.

There are three kinds of playlists Moodmixer can produce:

  • The non-stop running, endless playlist produced by the Mixer- music never stops
  • M3u playlists of a certain length with the playlist generator- playlist to be exported
  • scheduled playlists produced by the generator and started by the mixer.

The Mixer for non-stop play consists of a selection box for the actual title pool Moodmixer is allowed to select songs from (all songs, predefined collection or searchlist), five slider controls where you can select single values for each property or open a range of values for each property to narrow down selection and the set of rules mentioned above which can be activated.

The generator hosts the same setting-possibilities as the mixer but one can give the playlist a desired length or schedule it for a certain date and length.
You can set the Mixer to "play scheduled list” and any list will start at its designated date and time.

Of course a utility to rate/re-rate each songs while it is running is also implemented. Mass/batch-rating and also mass editing of Tags are included.

Moodmixer's smart algorithm already produces perfect results even if only a few titles have been rated! No need to do the whole job at once. It’s only getting even better!

The future:

Currently PWsoft is developing ActiveX components of the Mixing module (automated selection of songs according to personal ratings calculated by the smart algorithm), the search engine (quicksearch and fast,extensive search for combined criteria) and the speedy database- to be able to add Moodmixer’s auto-selection functionality to any other digital music software or hardware player. The mixing module with the smart algorithm is nearly ready.

You can download the latest Moodmixer version as winamp 2x/5.x plugin at http://www.moodmixer.com
More information, screenshots and detailed description can be found at http://www.moodmixer.com/Englisch/MMFunktionen.htm

Moodmixer is a registered Trademark and Copyrighted 2000-20004 by PWsoft

For inquiries please contact:

Pwsoft - Kerstin Laveatz
Mittelkamp 31
22043 Hamburg
Germany

Phone: 0049-40-760 80 725
Fax: 0049-40-760 80 727