Author Topic: Music Slicing (Read 5772 times)

Wander · « **on:** February 08, 2011, 05:37:07 am »

I've been trying to figure out a way to remove lyrics from a song. Anyone got any ideas?

WitchD0ctor · « **Reply #1 on:** February 08, 2011, 01:17:15 pm »

UHH, in real time using SFML? I dont think so, Maybe using an audio editing program like audacity, even then that might be a bit of a difficult task.

Wander · « **Reply #2 on:** February 08, 2011, 02:34:13 pm »

Haha. I figured it'd end up being something like that. Thanks.

l0calh05t · « **Reply #3 on:** February 08, 2011, 03:14:25 pm »

really removing the lyrics is almost impossible. The best option is probably to simply subtract the left channel from the right channel (as vocals are usually centered)

Irbis · « **Reply #4 on:** March 01, 2011, 12:43:35 pm »

Subtraction rarely works at least due to the compression.
You cannot remove vocals perfectly, but you can reduce them.
That can be done by using Fourier transform, reducing frequencies of the vocal range and performing inverse Fourier transform to recreate the wave back.

l0calh05t · « **Reply #5 on:** March 01, 2011, 03:10:21 pm »

Quote from: "Irbis"

Subtraction rarely works at least due to the compression.
You cannot remove vocals perfectly, but you can reduce them.
That can be done by using Fourier transform, reducing frequencies of the vocal range and performing inverse Fourier transform to recreate the wave back.

That will not work, as those pesky "vocal frequencies" are 90% of the important "music frequencies" as well

so this would reduce pretty much everything except extreme bass and extreme treble

Subtraction isn't perfect, due to - as you mentioned - compression, but also due to reverb, doubling and other stereo vocal effects, but it really is the best you can do. You could aditionally reduce the "presence" range of 2-3 kHz, but that will also affect your music quite a bit

In an ideal world you could use BSS (blind source separation) and vocal recognition to remove the vocal source. But in the real world... "that ain't happining"

devlin · « **Reply #6 on:** March 01, 2011, 05:52:36 pm »

You could also pay the author to make an instrumental version.

Irbis · « **Reply #7 on:** March 03, 2011, 09:48:33 pm »

Quote from: "l0calh05t"

That will not work, as those pesky "vocal frequencies" are 90% of the important "music frequencies" as well ;) so this would reduce pretty much everything except extreme bass and extreme treble

Subtraction isn't perfect, due to - as you mentioned - compression, but also due to reverb, doubling and other stereo vocal effects, but it really is the best you can do. You could aditionally reduce the "presence" range of 2-3 kHz, but that will also affect your music quite a bit

In an ideal world you could use BSS (blind source separation) and vocal recognition to remove the vocal source. But in the real world... "that ain't happining"

Of course, nothing really works for re-creating a minus-track, so I was just trying to give some hope;)

Quote from: "devlin"

You could also pay the author to make an instrumental version.

It is also possible to find midi tracks for most popular music, sound is not very "musical", but it's suitable for karaoke if that's the case.