Computer-Generated Sound Final Project Mu-Heng Yang - PowerPoint PPT Presentation

Computer-Generated Sound Final Project Mu-Heng Yang mhyang@Stanford.edu

Music Computer-Generated Sound Final Project Mu-Heng Yang mhyang@Stanford.edu

Algorithmic Composition Mu-Heng Yang mhyang@Stanford.edu

StkInstrument  Parameter tuning for timbre and volume balance  Melody Line (Rhodey)  Bowed (Chord Progression)  Bass Line (BandedWG)  Key Root (VoicForm)  Hihat (Shakers)  Bass Drum (BandedWG)  Snare (Shakers)

Composition Concept  Stable v.s. Random  Add some random variables in my software.  It decides when the notes are played and which notes to play  It decides when the drums are played and which drums to play

Song Structure  Song structure makes music different from sound  Verse  Chorus  Bridge  Chorus  Use random variable to make verse filled with less notes  Enjoy the music (https://ccrma.stanford.edu/~mhyang/220a/fp/fp.wav)

Machine Learning  Crawled online for thousands of songs in MIDI format  Process them into useful information  key, melody, chord, root, genre, instrument, artist, verse/chorus…  Develop algorithm to predict the next note given the current melody lin  Enjoy the music (https://ccrma.stanford.edu/~mhyang/220a/fp/fp-ml.wav)

Conclusion & Future Work  It now sounds more like human-composed melody instead of monkey randomly tapping keyboard.  It can easily scale-up or adjust to user-input  Compose with requested key, genre, instruments, artist style, song structure, and emotion.  I’ve implemented automatic accompaniment system using deep learning and would love to import it into this project!

Thanks!