I was wondering about the data - the starting notebook puts 44100 as the sample rate, however any check I do (with ffprobe for example) tells me that its sampling rate is in fact 48000. Was the 44100 a guess, or is that some insider knowledge and the saved .wav meta information are wrong?
This makes somewhat of a difference in compatibility with some packages, so clarification would be nice.
When I looked at the sample data it was all over the place: the initial voice corpus has it at 22khz mono while in the additional corpus it is either 44khz or 48khz and stereo (don't remember if all clips are stereo in the additional corpus or not)
I think that for the starter notebook, librosa offer the option of resampling the audio if you give a simple rate different than the default one. Also a sample rate of 44100 or 48000 doesn't make that difference.