Hi,
I'm trying to extract female vocal singer from a song. All I could understand about vocal extraction is that it's hardly possible to do anything but central channel cut, if the recording is mastered in certain way.
I did that, and I further applied low pass to mask unwanted content and also I pronounced the frequencies where vocal is spread, but I'm not happy with the results.
I then tried RX with Deconstruct module (which splits tonal and noisy contents) but still, result is not very good.
Going back to Audition, the track I'm working on, has a whole vocal range masked with (aggressive) harmonies from Turkish tambur. If I switch to spectral pitch view, a can clearly locate the voice, but in this view there is not possibilities for edits, like in spectrogram view. There is perhaps a reason why this is not possible, but I hope it's not some physical law.
So my question is - having the pitch information (let's say MIDI notes), is there any way I could apply this as a spectral filter?
What exactly are you trying to accomplish?
There's often an instrument playing the same exact note that's being sung, and then there are harmonics & overtones that overlap over a wide range.
I've never used RX, but it's supposed to have one of the best spectral editing tools. But some things are just impossible. Or, sometimes things can be done but the result is not "professional" sounding.
So my question is - having the pitch information (let's say MIDI notes)
If you could build a bandpass or band-reject filter that's tuned to MIDI, the results would sound weird! High amounts of data-compression (like what's used on cell phones) tends to filter-out everything except the most dominate frequencies, and it often sounds "weird" with the constant "re-tuning" of the filters.
I'm just trying to extract the vocals clearly as possible
So my question is - having the pitch information (let's say MIDI notes)
If you could build a bandpass or band-reject filter that's tuned to MIDI, the results would sound weird! High amounts of data-compression (like what's used on cell phones) tends to filter-out everything except the most dominate frequencies, and it often sounds "weird" with the constant "re-tuning" of the filters.
I couldn't have known that, and that is the exact result if I try to deconstruct tonal part at max value with RX.
Thanks for the useful replies, as always