Conference napalm::commusic_v1

    Studios do the "grab the <specific drum> off its track and use it
    as a MIDI trigger" all the time.  Pitch tracking is a bit harder,
    but possible on an isolated track for a monophonic line.  I think
    resolving chords on an isolated track is still beyond the state
    of the art, and pulling a single instrument (even if it's playing
    a monophonic line) out of a mix is also beyond the state of the
    art.  Amazing analyzers, these ear/brain things.
    
    len.
    

    Depending on what you're trying to snag, I'm inclined to agree with
    Len.  Take your bass line example, for instance.  Now let's suppose
    that you've got someone like Abe Laboriel doing his multi-pop thang, or
    a plethora of hammer-ons.  How you gonna track that? 

    Or take me and my stumbling, self-taught technique.  Even on a simple
    line, I doubt that you could capture and quantify (via MIDI) all the
    nuances that exist(ed).  Not that you'd want to, of course...

-b

    The legal question is interesting.  What would the DAT-police say about
    this?  There are clearly legal uses for such a technology (making
    scores from improvised performances, for example).  Put there is
    nothing to prevent taking a (copyrighted) record album of somebody and 
    running it thru this process.  Something like the uproar about sampling
    somebody else's performance.
    
    Dunno about bass lines.  It is hard to get the pitch right and also
    follow high-speed playing.  The problem is similar to what the
    photographer faces in dealing with the color temperature of a light
    source.  Our eyes/brains see sunlight on snow and tungsten light on
    paper as both being white, while camera film "knows" that the former is
    much bluer than the latter.
    
    How much would you be willing to pay for something that did this? :)

An interesting thought. I believe that more of the technology exists
today than we might expect, although it might be excessively complex
and expensive for a musical application.

I'd ask the department of defense. Most probably the
Navy. Signature analysis of acoustics has had quite a bit of
development over the years, especially in the area of submarine
warfare. In the 70's, I worked on a simulator of the S3A, anti-submarine
fighter. It did acoustic analysis, to identify submarine types, and
could differentiate many types of subs, wales, schools of shrimp etc..
Granted, that is a slightly different problem than decoding music.
But, if similar techniques were applied to music analysis, along with
the power of todays computers, I think it could be done. It's just a matter
of applying the right scientist to the job.

"Computer, please listen to this song, write me out the score, and
put the guitar parts in tab. Thank you, carriage return."

8^)
Rick

    Ah now, writing out the score is an Expert System problem, as it is not
    easy to tell the difference between a ritard and sloppy time keeping.
    You would have to tell it so much about the performance that you might
    as well do it by hand.  Some notation programs, notably "Finale" make
    some attempts at this.
    
    If all you want is a MIDI capture of a performance, that is just a
    signal processing problem.  MIDI is, after all, intended to convey the
    nuances of performance, not the formalities of western music notation.
    
    Not that both are not worthwhile endeavors in themselves.  But they are
    separate.

    re .5, specifically MIDI - well, not quite.  There are these
    troublesome MIDI clocks that are supposed to come 24 per quarter
    note, implicitly defining both the local tempo and the bar lines,
    were one to be able to intuit the applicable time signature.
    
    len.
    

    Those clocks are of no interest to anybody but a sequencer (or a drum
    machine).  They aren't even encoded in MIDI files.  All a MIDI file
    says is things like "n clock ticks after the previous event, thus and
    so happened".  The meaning of a "clock tick" can be anything you want.
    
    For example, many sequencers use 480 ticks to the quarter note.  A MIDI
    file tempo directive says how many *microseconds* per quarter note.
    So if you arbitarily set a tempo of 500000 (which is 120 quarter notes
    per minute), with a resolution of 480, you encode your delta times
    with 960 ticks to the second.  The actual event coding has no musically
    significant time information - just ticks.
    
    You also have the option (in the file header) to say that you are not
    using metrical time at all, but SMPTE time.  Then you can say things
    like "this file is coded for 30 frames per second, with a resolution of
    4 ticks per frame". This gives you 120 ticks per second.  (4 ticks per
    frame happens to be the MTC resolution.  Video SMPTE has 80 per frame.)
    
    So you can take a random series of events, musical or not, and encode
    them into a MIDI file in such a way that someone reading the file will
    be able to reproduce those events with the exact timing you wanted.
    Karl Moeller, for example, uses his sequencer as a tape machine, with
    no regard for what tempo the sequencer thinks it is using.  As long as
    your sequencer has enough resolution, you don't need to worry about it.
    (and all the good computer-based sequencers these days use 480 ticks
    per quarter note.)
    
    - Paul Dickson, MIDI file weenie.