Acapella

[MixMasterMax]

Is there some software that allows one to convert a song into acapella? I would like to keep just the vocals and remove as much of everything else as possible.

Thanks in advance

[BrunoRepublic]

In a word, no. There is no way to isolate different elements of a song without access to the original multitracks.

[MixMasterMax]

Thanks Graham_Start, that's what I suspected but needed to hear it from the experts.

[mynameismatt]

I've had success using the inversion method. Using cd--not vinyl :(

You need the song and the instrumental. Invert one of those using a program like Sound Forge. Inverting basically turns the waveform upside down. Copy the inverted track and paste it over the original in the EXACT same spot. What should result is everything not heard in the instrumental = the vocal.

I've been able to do it successfully with 2 songs. No, 100% of the background was not removed, more like 95%.

[MixMasterMax]

I am sorry mynameismatt. I have SF and tried inverting. I must not be doing something right because i noticed no change. Could you please try explaining your process in a different way. TIA

[Forrrce]

***minor log-in freakout - can't delete this***

[Forrrce]

Phase-shifting is relatively easy to do in the analogue domain, if you have a DJ mixer with a mic input and a mono button. If you can butcher and adapt some old leads, simply run a (cued) signal out of the headphone jack, with one channel looping back to a spare line-in; and the other channel going straight to a mic-in (always turn the mic volume down before doing this - and the signal must be centred, as opposed to panned either way). Switch the mixer's output signal to mono (this will not work if switching to mono affects your cued signal - it has to remain in stereo), then gradually bring in the mic volume to a desired level. Do the same with the line-in's signal and you will gradually skew the phase.

Use a *very* stereo record or CD's signal to test this - spoiling the phase in this way will minimise the mono signals in a recording and leave the radically panned stereo signals intact, all in one mono signal. So-called voice removal programs work on exactly the same principle and unless you're dealing with well-preserved '60s recordings where the engineers were stereo-happy and panned all instruments to a single channel and left the vocals in the middle, this technique is largely ineffective and no more than a gimmick. Problems inherent in recording, as well as sound effects you may not notice normally,will surface (and can even dominate) when you shift the phase - like print-through, various leaks and badly degraded source material. Having said that, I once gave Bob Dylan's 'Like A Rolling Stone' the treatment from an old cassette copy and I completely removed his vocals, leaving the music intact. I've never been able to do this as successfully since, even with the CD.

On the plus side, phase-shifting is very handy for straightening out some poorly-mastered records and CDs - and I even used to align tape-deck heads using this method.

[BrunoRepublic]

Originally Posted by mynameismatt

I've had success using the inversion method. Using cd--not vinyl :(

You need the song and the instrumental. Invert one of those using a program like Sound Forge. Inverting basically turns the waveform upside down. Copy the inverted track and paste it over the original in the EXACT same spot. What should result is everything not heard in the instrumental = the vocal.

I've been able to do it successfully with 2 songs. No, 100% of the background was not removed, more like 95%.

I've thought about trying this before, but never went through with it because:

- I don't have a lot of songs on CD with vocal/instrumental mixes that are otherwise identical.
- The source material would need to be digital, as even the smallest pitch variations on analog tape will throw everything off. The two tracks would need to be lined up to the exact sample for this to work at all.
- I would imagine that the vocal would affect the dynamics at the mastering stage, so the actual levels of the instrumentation would be slightly different at different points in the mix... thus the process would be rather ineffective.

But I do have a couple of songs that I think fit the criteria, so I'll give this a try later.

[deepdjdanny]

there was a software plugin called Pandora ? for SonicWorx? Mac... that claimed it could pull whatever elements out of complex audio using neural network algorithms (?) german (?)
I heard the examples & they were good... acappella made in about 4 passes...
The plugin probably never made it to market, got aimed at Forensics to clean up speech in telephone calls, or spy stuff...
I'd pay money for this software!!!!
turn it into an external fx unit for dj mixer....

[Bernie]

Apparently development on it has ceased. See http://products.prosoniq.com/cgi-bin...detail&refno=9