wammie wrote: runagate wrote:
heuermh: LiCK Library for ChucK? Thanks for this. I've got ChucK on my list of things to investigate more closely, since I was trying to make a pre-anylized FFT file format for audio and suddenly had the bright idea to search Google for those terms and found TAPESTREA ( http://taps.cs.princeton.edu/
Holy crap, that TAPASTREA program looks amazing!
Yeah, it blew my mind. I simply cannot believe it already has been in existence for years.
It will save me about a bajillion hours of development.
I intend to batch analyze and then auto-group any audio files by hash tags according to averaging of spectral similarities, with certain pre-defined categories sifted out individually (hit-hats, inharmonic bells, etc.), and then then create a multi-dimensional matrix of comparisons for velocity, root pitch and "attack style" (say, a picked vs. a fingered vs. a slapped electric bass) and for each of those interstices provide a change-over-time profile for the pitch, amplitude, expression (vibrato and tremelo, initially), attack transients and spectral (harmonic overtones, non-harmonic partials and relatively-invariant formant) envelopes.
What I'm working on is a universal networking protocol for control data for music, so it needs to have certain (several thousand, at this point) composition- and performance-oriented specifics defined in frameworks way, way at the beginning to make sure they're extensible further along the trail when actual APIs exist to be communicated with
so it's easily extensible. So that described above is one part of it which will allow for vastly more expressive higher-order organizational UI patches for multi-dimensional interface control (read: gyroscopic, multi-touch, AR) that doesn't really exist now to enbable that those extra mdimensions of expression to, well, control something useful, in this case extremely easy to use but complex and correlated envelopes. MIDI and even OSC doesn't even remotely allow for this, especially since it lacks hierarchies of complexity perspective that maxes out with higher-order perspectives for ease of use in addition to extreme complexity (nowadays it's really an either or proposition). This should work really nicely with additive/spectral, physically modelled and pre-analyzed FFT'd (time and pitch domain, both) sample synthesis.
The really startling part of it, to me anyways, is that this should allow taking several actually existing data points on this matrix of parameters and to interpolate and fill in the zillion other variations of sound based on the extracted definitions... probably not in realtime, but batch-processed, and allow those parameters in the matrix to be user edited and thus allow for a completely new kind of sound design. Too bad I only have one programmer partner
I can still barely believe that TAPESTREA will allow me such a head start. I can't believe I stumbled on it by googling my own, at the time,completely unimplemented idea and someone far smarter than I made a program a few years ago that's that powerful. And, thankfully, Open Souce.