|
From: Philipp M. <the...@go...> - 2018-06-01 08:03:45
|
Hey everyone, Could someone explain to me the meaning behind the so-called *swing* factor in the humanizer? In line 1281 of ///src/core/src/hydrogen.cpp/ it is implemented as a *constant shift* forward (backward? haven't had time to delve into the audio engine that much yet) in time. I would like to implement some additional sources of noise in the humanizer. I don't really see how the little errors we human make a completely independent and could we described by Gaussian white noise. Instead I want to introduce - Pink noise, since analysis of fluctuations in music claim this to be the most likely type of noise since the 80s http://mlab.uiah.fi/~eye/mediaculture/noise.html - State dependent noise. The variance of the generated noise should in my opinion (but I'm not a drummer myself, please correct me if I'm wrong :)) dependent of the velocity of the hit. Or to put it simple: the stronger you hit the drums, the bigger the fluctuations of the amplitudes. - Long-range correlated noise. There have been some new findings by indicating that long-range correlated noise is indeed the best to describe the fluctuations in rhythms. https://www.researchgate.net/profile/Holger_Hennig3/publication/257020003_Musical_rhythms_The_science_of_being_slightly_off/links/56b3a0f408ae1f8aa45350dc/Musical-rhythms-The-science-of-being-slightly-off.pdf I wrote some mails to the main author Holger Hennig since there is a patent on the algorithm. But for a free and open source project like Hydrogen there is nothing to worry and he is happy if we spread the word of correlations. But before building on top of the current humanizer I would like to understand all its components. Cheers, Phil |