zero reducing vocals ?
are there software that could reduce the vocals of a song to an unaudiable level -- without reducing the audio quality? i have one called Goldwave, but it cant actually do that. When i reduce voclas with it, i still could here the voice as loud wispers, and the audio quality drops. if i reduce it more, quality becomes crap. any recommendations? thx.
Well, unless you have original recordings on a multichannel media, you can't separate voice from other sounds on the recording - all of the sounds are mixed together into the two stereo channels (obviously, if all of the vocals are on the left/right channel, and the instruments are on the right channel, yes, it would be simple).
Any software that attempts to do this will use some "intelligent filtering", but it's never going to be doing EXACTLY what you want - it will at best guess right most of the time.
Tell me, if you have a waveform (of an audio track) on the screen, how do you know what part of that is the vocals, and what is for example drums or guitar?
ok. but i guess it might be easier to filter out the dialogs from a movie (at least for some extent) and to bring out the underlying soundtrack (vocals are more infrequent and clearer than in songs)? anyone know of any software designed to do this?
With advanced software it's possible to separate any sound source from the remaining sources. The quality of the results may vary and it involves a lot of work an patience on behalf of the person using the software.
It is however necessary that other sound sources don't superimpose themselves on the source you are trying to extract. As long as there is only the faintest audible representation of a sound source, it can theoretically be extracted and manipulated, no matter the number of channels.
Good luck finding the money to buy these.
EDIT: Was forgetting the link... http://www.cerlsoundgroup.org/Loris/ for an advanced open source alternative
Most karaoke machines work by assuming that the vocals are at stereo center, and the instruments are always displaced, at least somewhat, from center. Therefore if you correlate the left and right channels you get a strong signal which approximates the vocal track, and you can subtract that out. In practice, the correlation might not even be necessary if the vocals are strongly to the center and everything else is strongly to the sides:
Originally Posted by matsp
L = l(t) + v(t)
R = r(t) + v(t)
Then the difference is:
L - R = l(t) - r(t)
And the v(t) has disappeared.
yeah, vocal reducing software works better if the vocals are flat and has no resonance in either channel.