Freesound Forums

I got such brilliant feedback from the community for my last questions. Now I have another.

I want to take a voice recording - or re-record it if necessary - and bring it back in time. I want to make the interviewee sound like he sounded when he was a kid. Any suggestions on how to get that effect - to transmute a man's voice to a kid's? I want people to recognize it as the same voice.

All ideas are welcome.

Some sort of voice pitch shifter with formant control. But I'm not sure about the quality. Steniberg had a little plugin that could do the job, Also Waves could do.

Thanks. I'm not an expert, though. If you could be a little more specific, that'd be great:
Does 'pitch shift' simply mean making the pitch higher? When I do that it just sounds kind of camp, like a mock-gay voice (no offense to anyone). Formant control?

Can you remember what the steinberg plugin is called?

Also, what's Waves?

Sorry for the bother, but I'm new to this.

Thanks.

This (and other plugins of this kind) should be along the line of what ayamahambho is suggesting:

http://www.g200kg.com/en/software/rovee.html

You might want to watch the video on the page, to get a better idea of how it works.

HTH

This stuff is quite complicated. The human voice is basically a vibrating device (vocal chords are like a string instrument) which vibrates at some frequency depending on the sound the speaker is making. This creates a fundamental frequency and a set of harmonics (or overtones) at integer multiples of the fundamental, just like a string instrument.

Added to this is the vocal tract which is essentially a resonant cavity - a bit like a wind instrument. The vibrations of the vocal chords excite a number of resonant sounds in the vocal tract. The frequencies of these resonances depend on the length and shape, which vary as you move your jaw, lips, tongue, soft palate etc. For vowel sounds you see three or four distinctive peaks in the frequency graph (between about 300Hz and 3000Hz for men), and these are what give the vowels their characteristic sounds (and are specific to each speaker). These are the formant frequencies.

As you say, if you just shift all the frequencies up using naive pitch shifting, you stray away from a human voice sound. A better way is to change the fundamental frequency and the formant frequencies independently which is what some software allows you to do. Even this is some way from perfect because the formants vary between male and female voices. To do it properly, you really need to simulate the resonances that form in a different shaped vocal tract, which is non-trivial to say the least.

In addition to ayamahambho's suggestions you could have a look at Rovee voice changer (I think they may even do a free version) and MorphVOX Pro. At the top end there are things like Melodyne. I'm not sure that any of those will convincingly achieve what you want though.

Hope that helps.

I can see I have a lot to learn. Every post here has been really helpful. I appreciate all the feedback. Now I have to go study it all.

most films would simply dub the younger voice than opt for any hint of processing.

most films would simply dub the younger voice than opt for any hint of processing.

Hey Martian,

Having spent a lot of time looking at this I believe yuou're right. Just getting the speaker to do a kid's accent is way easier!

okay...

There is a cute software called MorphVox pro, which works with audacity and many audio editors.

http://www.screamingbee.com/product/MorphVOX.aspx

but I prefer rovee.

if you want to keep it simple

i had to create a kid-like voice in a game where i did the audio. i found pitch shifting through granular synthesis or similar models too disturbing for the quality of the voice. my solution: let the speaker speak slowly (not really slow but just take the time) and then replay the file a bit faster (i.e 3 semitones or 1.25 speed) maybe take a bit of the low frequencies off too.

Best option for me: record a real kid.

Ok, Thank you..