This is a small test dataset for Automatic Speech Recognition in Darija language, built from 2 captioned videos of the WikiTongues project:
nawal
anass
Process:
each webm video has been converted to monochannel 16khz wav files with ffmpeg :
ffmpeg -i WIKITONGUES-_Nawal_speaking_Moroccan_Arabic.webm.1080p.vp9.webm -ar 16000 -ac 1 nawal.wav