• The goals of this project
  • The tools used in this project
  • Tools invented
    • steady-sound.lisp (by Tak)
    a sample lisp code to pick up a frame of spectrum and make a sound all of whose frames are made the same as that particular frame in the original.
    presentation
    loops.lisp (by Harvey)
    a lisp code which creates a looped version of the original sound.
    express.lisp (by Tak and Harvey)
    ATS functions for superimposing vocal-like expressiveness on ATS-formats sounds.
    1) "vibrato-sound" is a multifaceted tool designed to give vocal-like expression to an ATS sound. Vibrato and amplitude envelopes are applied over the duration. That is, the vibrato is a time varying effect.
    2) "transfre-specenv" allows you to do some frame-wise modifications, like mophing.
    morphing example: takuya-a-i-transient.snd
    Through the obsrevation of actual diphones, it proved that the formant behaviors at vowel transients are very complicated. So I gave up the idea to imitate them and simply tried a formant cross-fade in this example.

  • Accessaries invented
    • wave2lisp (by Tak)
    a dirty shell script for file conversion automation: It automatically issues commands to convert a sound file to an analyzed data file. Note that this works only for sound files prepared for this project which are under specific naming convention.
    idlenext (by Tak)
    another dirty shell script to find an idling black NeXT around here. Er wird den hoeren, ,,Wachet Auf! Schlafet Nicht! Werket Hard!
    playit.lisp (by Tak and Harvey)
    a tiny lisp function to hear an analyzed data object. It also applies optimizing function if the object is not optimized yet.
    fins.clm (by harvey)
    a non-real-time vocal instrument which includes the envelope-controlled vibrato-rate & -depth, structured by 3 formants with information of center frequencies and Qs.


  • Procedure

    • Diagram (by Ching)

    1) Several English vowels are recorded.

    ** All of the sound sources of this project are located in '/usr/ccrma/snd/fujishim/mus220b/' directory.

    2) Some of the recorded vowels are looped in the wave domain by using SoundDesigner. (by Unjung and Ching) They are the followings:

    Female: /a/ in pitch 'a'(440hz) l-uaa-mix3.aiff
    /a/ in pitch 'd'(293hz) l-uda-mix3.aiff
    /i/ in pitch 'a'(440hz) l-cai-mix3.aiff
    /i/ in pitch 'd'(293hz) l-cdi-mix3.aiff
    Male: /a/ in pitch 'c'(131hz) l-tca-mix3.aiff
    /a/ in pitch 'g'(175hz) l-tga-mix3.aiff
    /i/ in pitch 'c'(131hz) l-hci-mix3.aiff
    /i/ in pitch 'g'(175hz) l-hgi-mix3.aiff

    2) smsAnal is used in 'wave2lisp' in order to analyze the original sound. Here are some explanations on the parameters. smsflags.

    3) smsToATS is used to transfer .sms files to .lisp files in order to resynthesize sounds or do the transformations of them in ATS environment

  • Steps

    • Perspectives: