The Music Video Generator produces a computer-directed film in response to audio files using Max/MSP and Jitter. It uses Tristan Jehan's 'analyzer~' patch to map audio features to video parameters such as color, clip selection, and crossfade. The user's possible aesthetic choices are constrained by its interface, which allows for the selection of an audio file to analyze, six film clips to use, and manipulation of color mapping. Randomness in parts of its process ensures that the patch generates a different film each time it runs.
The film jumps between clips when attacks in the audio are detected by 'analyzer~'. Each attack produces random numbers that are used to select and layer two clips. Color manipulation in the film is based on loudness in different frequency bands. This often produces an eerie, murky start to the films since many songs begin with less spectral range, and loudness in only one frequency band outputs mostly dark, solid colors. As the song progresses and the spectrum is filled out, the clips often become clearer and easier to decipher. The generator is most effective with music that has varied loudness throughout. Overall loudness of the track is mapped to crossfade between the two layered clips, meaning that when the audio is significantly louder or softer, one clip emerges more clearly from the murkiness. An interesting effect can be produced by setting all six of the video inputs to the same clip: this layers different time points of the same clip on top of itself, and the film skips through time with glitchy color transformations.