Content-Based MP3 Playlist Generator & Jukebox
What is VisiTunes?
VisiTunes is a content-based MP3 playlist generator and jukebox. It uses the audio content of MP3 files to calculate a measure of similarity between songs, and uses this information to generate and play lists of songs that it determines to be similar to a seed song selected by the user. It can create playlist files in M3U format that can be played with any media player program, or it can be used as an automated jukebox, selecting and playing similar song files. View the help file.
How would I use VisiTunes?
VisiTunes can be a great help if you want to select songs from your collection that match a certain style or type of music. It can perform this task automatically, saving the time and manual effort of searching for the right songs. You can use VisiTunes to select songs to copy to your MP3 player or PDA, to select songs to burn to a CD, or just pick a song and let VisiTunes select and play similar songs on the fly in jukebox mode. Of course once VisiTunes makes its recommendations, you can still edit and fine tune the list to your liking. VisiTunes helps you dig deep into your music collection, bringing up songs you may have forgotten about, or making connections that might not occur to you.
Can my computer run VisiTunes?
The current version of VisiTunes runs only under Windows (any version from 98 through XP) and works only with MP3 files. If you have a PC running Windows and a music library in the MP3 format, you can probably run VisiTunes. Future versions may support other file formats and other operating systems. VisiTunes continues to be a work in progress, but the current full featured pre-release version is available for free download and unlimited use.
How do I install VisiTunes?
Download and execute the installation file <here> and follow the on screen instructions to complete the installation. See the help file for instructions on how to set up your MP3 library with VisiTunes.
VisiTunes running in auto-play jukebox mode
Setting playlist parameters
What about other playlist generators and mix makers?
Other playlist programs rely on manual clasifications submitted by experts or users to determine similarity between songs. Humans manually categorize songs as fast or slow, happy or sad, and into pre-existing genres, or look for the existence of predefined attributes. In contrast, VisiTunes relies on the idea that if there is a music genome to be discovered, it will be discovered in the data of the audio file and not through external observation.
VisiTunes is the first application available that performs audio similarity analysis functions automatically and with no user intervention, and in a free-form manner, with no pre-existing categories or attributes. The only criteria used in the analysis is the sound data actually contained within the song. This means that VisiTunes can work work with any songs in your library no matter how obscure, since it is not dependent on prior manual classification. And the connections suggested by VisiTunes are based on reality, not biased by an individual's preconceived ideas about genres or artists.
Does this content-based analysis actually work?
In some ways it works very well. The instruments in the music we hear usually produce sounds within a narrow spectrum. An acoustic piano will produce very different frequencies than an electric bass, or a saxophone, or an analog drum machine. VisiTunes groups songs by the presence of sounds in these frequency ranges, so a group of solo piano pieces will be seen as very similar, and a group of songs with prominent electric guitars will be seen as similar. There are cases where the results may be surprising to the listener. One case would be very different types of instrumentation that share common frequency ranges, such as an uptempo horn-driven big band song being seen as similar to a hard rock song. Another would be songs with similar types of instrumentation but from artists considered to be working in different genres due to geography, ethnicity, time period, costume or target audience. Your satisfaction with VisiTunes is likely to depend on your attitude toward these surprises.
View some sample playlists generated by VisiTunes:
Select the links below to see playlists generated by VisiTunes for these songs.
Urge Overkill - Positive Bleeding (hard rock)
Pete Seeger - If I Had a Hammer (folk)
Dr. Dre - Nuthin' But a G Thang (hip hop)
Beethoven - Fur Elise (classical)
Meet the author:
VisiTunes is being developed by Scott McCaulay. Scott is a PhD student in Information Science at Indiana University with research interests in Music Information Retrieval (MIR), content-based similarity analysis and machine learning. VisiTunes is an attempt to take the concepts being developed in MIR and simplify them to the point that they can be implemented in a real world application. Scott's home page is http://www.mccaulay.net
Music Information Retrieval (MIR) links:
http://www.music-ir.org - Virtual home of music information retrieval research
http://mirsystems.info - A Survey of Music Information Retrieval Systems
http://www.ismir.net - The International Symposium on Music Information Retrieval
Send comments, feedback, suggestions - send email to VisiTunes