5

Suppose I have a backing track for a song and then record two different singers each performing the lead vocal for the song.

Is any signal processing possible that would make it sound like the first singer was physically becoming the second singer? So not a simple cross-fade between the two performances, because you can hear there are two people singing.

  • How convincing would that be if we were transitioning from a male bass to a female soprano, complete with voice range changes? – Dekkadeci Feb 27 '18 at 16:07
  • @Dekkadeci If by 'convincing' you mean 'natural', you'd have to make the notes in the composition correlate somewhat with the formant of the voice, so in your example you'd need to compose a general upward pitch gradient in the vocal line to match the increase in the (possibly synthesized) 'resonances' of the morphing voice. – topo Reinstate Monica Feb 27 '18 at 18:48
  • I think there are sampling synthesizer plugins that can morph from one sample to another. So you could try that route. I would expect it to sound a little (or a lot) strange during the transition. – Todd Wilcox Feb 27 '18 at 19:36
1

A cool trick I’ve seen is to extend the last note to where the other singer hopes on and continues singing, it’s a seamless smooth transition. It sounds amazing when done right,

1

This would be my go-to tool for any type of audio morphing, especially voice. https://kyma.symbolicsound.com/sound-and-video-clips/ It is a fairly easy to learn GUI system that is also extendable with code. We had the Kyma/Capybara system in one of our college electro-acoustics laboratories, as it is capable of many complex DSP algorithms.
Essentially there are timeline views of the chronological performance and filter windows where you may construct your sound morphs graphically using icons.

0

The human ear is very good at differentiating sounds, so without noticeably altering the vocal tracks you will have a hard time doing this.

One of the recognizable differences in the voice is the Timbre, being made up of the frequency spectrum and the sound envelope that lets us identify who we are listening to.

You could possibly use frequency spectrum analysis to identify the differential between the two voices and create a frequency filter that you could could apply to each voice to make them sound more alike. An EQ filter may also need to be applied.

You would then either apply the filtering to both vocal tracks allowing the cross-fade to be less distinct, or apply the filtering as an effect before or during the transition to make the voices indistinct while you switch them.

I'm not sure what software you would use, but I'm sure the capability exists.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.