On Saturday 13 February 2016 16:53:29 Mauricio Kanada wrote:
> I was thinking in a new sample format based on spectrum and not on
> waveform. The idea is create samples that sounds good without the need of
> thousands of samples. Imagine a piano library with only 5 velocity layers
> stored in spectra format (maybe in mdct format) and the engine blend the
> spectrum at run time. Casio has this technology and name it as
> 'Multi-Dimensional Morphing'. Kontakt name it as AET (Authentic Expression
> Technology), but as I know, it is at edition time, not at runtime.
Well, this idea is not new and is already used in products for years. It was
also discussed here couple years ago when we started to work on additional
sampler engines/formats. Keep in mind that performing things in the frequency
domain can be quite expensive. That's why it is mostly still more effective
overall (for the musician) to play with a huge set of samples in the signal
domain instead.
And yes, sure would be nice to have, however so far simply nobody has yet
started working on this with LinuxSampler. Since this is essentially a spare
time project for all of us, it is all a matter of personal priorities.
CU
Christian
|
