Audio difference extraction -Sound Forge - PC - MAGIX Community Forums Archive

Subject:Audio difference extraction Posted by: seifer_almasy63 Date:2/19/2003 4:39:36 PM Hey, I'm a video producer but I use SoundForge to treat all my audio. Basically I'm trying to remove the music from a piece to camera I did a little while ago. I want to use the voice to a piece to camera but I don't have the original recording it's been lost somewhere over time. But I have the original music that was laid over the voice and I have all the levels and everything that the audio was set at. Basically I was thinking does anyone know if it’s possible to perform an audio-difference extraction so I could tell it to remove the music? Thanks very much
Subject:RE: Audio difference extraction Reply by: Chienworks Date:2/19/2003 10:10:44 PM Well, in SoundForge you could always invert the music, then paste/mix it with the combined recording. You can even adjust the relative mixing levels of the two files to get the best cancellation. Your two big problems will be sync/speed and distortion. There's no guarantee that the camera recording will be the same speed as the original material, and even if it is, getting them to paste "in phase" will be tricky. If you're out of phase by even a sample you'll compromise the subtraction. Also, whatever the music when through to be re-recorded could introduce it's own changes & distortion that will affect how well this works.
Subject:RE: Audio difference extraction Reply by: seifer_almasy63 Date:2/20/2003 5:43:18 AM Ok thanks for the help, the audio shouldn't have gone through any changes. I mastered it to digi betacam so it should be identical. I think I might be able to do it now because I don't need to competley cancel the audio out of the peice to camera because I'm going to put another one in its place which will hopefully hide it. Thanks for the help
Subject:RE: Audio difference extraction Reply by: Geoff_Wood Date:2/20/2003 3:26:30 PM The copy will not be sample-oramplitude-accurate to the original, and may drift significantly over the whole length. I suggest you time-stretch the copy to make it equal at two signficant identidiable points, and normalise. geoff

Subject:Audio difference extraction

Date:2/19/2003 4:39:36 PM

Hey, I'm a video producer but I use SoundForge to treat all my audio. Basically I'm trying to remove the music from a piece to camera I did a little while ago. I want to use the voice to a piece to camera but I don't have the original recording it's been lost somewhere over time. But I have the original music that was laid over the voice and I have all the levels and everything that the audio was set at. Basically I was thinking does anyone know if it’s possible to perform an audio-difference extraction so I could tell it to remove the music?

Thanks very much

Subject:RE: Audio difference extraction

Reply by: Chienworks

Date:2/19/2003 10:10:44 PM

Well, in SoundForge you could always invert the music, then paste/mix it with the combined recording. You can even adjust the relative mixing levels of the two files to get the best cancellation. Your two big problems will be sync/speed and distortion. There's no guarantee that the camera recording will be the same speed as the original material, and even if it is, getting them to paste "in phase" will be tricky. If you're out of phase by even a sample you'll compromise the subtraction. Also, whatever the music when through to be re-recorded could introduce it's own changes & distortion that will affect how well this works.

Subject:RE: Audio difference extraction

Reply by: seifer_almasy63

Date:2/20/2003 5:43:18 AM

Ok thanks for the help, the audio shouldn't have gone through any changes. I mastered it to digi betacam so it should be identical. I think I might be able to do it now because I don't need to competley cancel the audio out of the peice to camera because I'm going to put another one in its place which will hopefully hide it.

Thanks for the help

Subject:RE: Audio difference extraction

Reply by: Geoff_Wood

Date:2/20/2003 3:26:30 PM

The copy will not be sample-oramplitude-accurate to the original, and may drift significantly over the whole length.

I suggest you time-stretch the copy to make it equal at two signficant identidiable points, and normalise.

geoff