# Clean podcast audio

Split using VAD -> detect at least 30% cleaned speech, ~352 hours, https://github.com/huseinzol05/malaya-speech/tree/master/data/podcast