Mp3 player/music editor that deletes spoken words and plays music

October 26, 2021 8:41 PM

Is anyone aware of a podcast/mp3 player that will automatically take an mp3 file and then either skip/ignore (or delete) spoken-word/conversation in that mp3 and only play the music. Or, alternatively, a sound editor that could convert the file to a “just music” mp3. The player must work on android. The editor could work on either android or MacOS.

I have some music podcasts downloaded that have great selections of music. Over time, the increasingly banal conversation of the hosts, as well as news and weather and ads inserted into the program annoy me. I want to cut the chatter and play the music.It doesn’t have to be perfect, just get rid of most of the talk.

posted by lalochezia to Computers & Internet (4 answers total)

Well, you could load the mp3 into Audacity and just snip out and save the songs as individual mp3s (or just snip out the conversation and save the shortened episode as a music-only mp3).

If you’re looking for something that does this in an automated fashion I’m not sure you’ll find anything.
posted by sesquipedalia at 9:03 PM on October 26, 2021

Reminds me of this. If you want this to happen automatically, you are asking the impossible.
posted by chuntered inelegantly from a sedentary position at 10:30 PM on October 26, 2021

Easy, take an FFT over the time and look for the sections that match the frequency range of human speech and cut those out. There are de-vocalization sorts of plugins that can *mostly* take the humans out of a piece of music. Finding an app pre-built for just that (or editor with that feature) might be the hard part. Cutting out monologues and news reports and weather mixed inbetween the music things shouldn't be much of a problem.
posted by zengargoyle at 12:14 AM on October 27, 2021

Almost every music player will handle playlists in CUE format [1] [2] which allow you to specify talk as hidden items before the music. Finding the talk is a challenge I'm looking at for a personal project and Zengargoyle is right that you should be able to see the difference between full-frequency music and limited frequency ranges of human speech, so a visualisation will help mark the start/stop times of the music. I don't have an automated solution for that, but I'm experimenting in the space.

1: https://en.wikipedia.org/wiki/Cue_sheet_(computing)
2: https://wiki.hydrogenaud.io/index.php?title=Cue_sheet
posted by k3ninho at 4:35 AM on October 27, 2021