Good voice morphing tool?

[alexis]

Howdy!

I was wondering if there's a tool that I could play a self- recorded vocal line to, and then play an album cut with another voice (not having access to the isolated voice track from the album), and then ask the tool to give me certain aspects of that album voice (sandpaper like Rod Stewart on Maggie May, for example)?

Failing that, if I built a library of others' isolated voices (online live and AI singers) could something like that get done?

Thanks!

[Drew Stephenson]

Paul White reviewed one in last month's mag I think. Will go check in a bit.

[alexis]

Paul White reviewed one in last month's mag I think. Will go check in a bit.

Wow, thanks!

For Reference:
https://www.soundonsound.com/reviews/dr ... s-vocoflex
https://youtu.be/9sJ7bghcraQ?si=9xUXiMDK2whoLC2d

The examples on the creator's own website weren't that inspiring IMO - the amount of morphing from one voice to another was not as much as I'd hoped, and even so there was a noticeable synthetic quality that I'd hope to avoid.

I guess Rod Stewart is safe from his voice being abused in my creative works!

I don't know whether "blending" an AI voice that has the characteristics I'm looking for, 6 or 12 dB under the lead voice, would do the trick. Worth a try perhaps!

Thanks again, Drew

[James Perrett]

even so there was a noticeable synthetic quality that I'd hope to avoid.

The original sounded synthetic to me - as if it was AI generated.

[amanise]

I think that for a while anyway, these kind of things are likely to be a bit disappointing because of the way we latch on to the human voice and instantly detect flaws or unexpected features.

That said, there are a couple of tricks which can be done. I'm sure these are very old hat and you already know them. There'a a reasonably good tool called 'Choir' from Antares that adds synthetic copies of the original voice and fiddles a bit with key aspects of the original to make them sound different. It was free when I got it, but it might now be part of the Autotune suite. If you mix a bit of that in it can add a bit of spice. Not too much.

Same with distortions. You can run a take through a distortion device of your choosing and onto a new track, and then use a volume envelope to bring a few dBs of it in here and there in the background so you know its there but can't tell what it is. Flangers also work well like this. Part of the trick would be knowing which techniques the original producers used to get the 'Rod' sound. It's likely to be a specific aspect of his voice that a great producer brought out with a mega expensive compressor or something. Find out what that was and Rod's your uncle!

But it's not 'give me Rod singing my tune'. Yet. I'm sure it will be, pretty soon. There may already be something close from Melodyne. I thought I saw a video from them describing virtual AI singers you can have voicing your takes for you. Unless it was another Antares module. Can't remember. It's all a ghastly nightmare.

[Drew Stephenson]

https://www.soundonsound.com/reviews/dr ... s-vocoflex

That was the one I was thinking of.

[N i g e l]

I think that for a while anyway, these kind of things are likely to be a bit disappointing because of the way we latch on to the human voice and instantly detect flaws or unexpected features.

very true. Cakewalk used to have the Roland V-Vocal which allowed elastic audio, pitch and formant shifting of a recorded voice. Only minor changes* were realistic, as humans are highly evolved for speech and listening to speech.

* depending how much else is going on in the mix to act as a concealer.

[alexis]

... Cakewalk used to have the Roland V-Vocal which allowed elastic audio, pitch and formant shifting of a recorded voice. Only minor changes* were realistic, as humans are highly evolved for speech and listening to speech.

* depending how much else is going on in the mix to act as a concealer.

I used to have a TC Helicon VoiceOne hardware box about 20 years ago, great promise, I remember working so hard and still being disappointed, only minimal changes to the voice were acceptable (it did a much better job at doubling though!).

The technology doesn't exist now apparently, and certainly didn't then for transparent voice processing of that sort to any significant degree.

[N i g e l]

another angle is Solaria and her mates on Synthesizer V.

From an audio track Synth V can analyse the singing into pitch & phonemes for resynthesis by one of the synthetic voices.

I think I watched a uTube on that and was reminded of one of the characters from southpark.

An old trick from multitracker days is to record with varispeed pitched up or down. When you playback at normal speed the voice formant will shift as well as the pitch. Its a handy technnique but no Rod Stewart gravel !

Rod Stewart & gravel (and pot holes )
https://www.youtube.com/watch?v=-P836SYVIBI

[alexis]

...

An old trick from multitracker days is to record with varispeed pitched up or down. When you playback at normal speed the voice formant will shift as well as the pitch. Its a handy technnique but no Rod Stewart gravel !

Still doing that, Revoice Pro, great for doubles, etc.

Rod Stewart, David Ruffin, c. 1963/4 John Lennon ... I wonder whether it's technology limitations vs lack of widespread interest/market that is keeping a plug in from being developed that applies those sorts of gravel effects to a vocal
.
.

Rod Stewart & gravel (and pot holes )
https://www.youtube.com/watch?v=-P836SYVIBI

"My Ferrari can't go through here at all!"

[amanise]

The 'old school' ways of bringing out gravel and wotnot are something I'd love to learn more about. I occasionally manage to get a gravelly sort of thing going on - but it's always there by accident, not design. Though it probably relies of bits of gear I could never afford. These things appear to be closely guarded secrets

[alexis]

The 'old school' ways of bringing out gravel and wotnot are something I'd love to learn more about. I occasionally manage to get a gravelly sort of thing going on - but it's always there by accident, not design. Though it probably relies of bits of gear I could never afford. These things appear to be closely guarded secrets

I think these were just voices (Rod Stewart Small Faces and Maggie May era, David Ruffin of the Temptations, Lennon c.'62-'64 for example) that were simply* naturally gravelly, smooth as #24 sandpaper , no need for studio wizardry/ bits of gear. I base this on how these guys sounded in live shows, where engineering was at a minimum (esp back then), live studio takes (from fly on the wall YouTubes), and the like.

Loud Kurt Cobain I think had the same voice, not sure about that. The singer from Jet, IIRC, also, etc.

*Not simply, actually ... what they all had in common IMO wasn't just gravel/sandpaper, but also the ability to sing like that on pitch somehow!

[amanise]

The 'old school' ways of bringing out gravel and wotnot ...

I base this on how these guys sounded in live shows, where engineering was at a minimum (esp back then), live studio takes (from fly on the wall YouTubes), and the like.

.... The singer from Jet, ....

*Not simply, actually ... what they all had in common IMO wasn't just gravel/sandpaper, but also the ability to sing like that on pitch somehow!

There was/is a Brit singer called Paul Young back in the 80s, who had the most versatile voice imaginable and could do pretty much anything he wanted with it but was a nightmare for FOH because it was so quiet. According to his bassist they always had to have the main mic really high compared to everything else, which was a pain if anyone got the wrong one.

Jet were an odd band. One album then nothing. I saw them at Brixton. Their keyboard guy was using a couple of laptops. Not bad live though, fairly faithful to their album.

I used to think you should never do anything in the studio you couldn't reproduce live. But not any more. All that puritanical stuff went out of the window years ago

[N i g e l]

There was/is a Brit singer called Paul Young back in the 80s, who had the most versatile voice imaginable and could do pretty much anything he wanted with it but was a nightmare for FOH because it was so quiet. According to his bassist they always had to have the main mic really high compared to everything else, which was a pain if anyone got the wrong one.

Paul Young was very big, hes the 1st one to sing on Band aid 1, followed by Boy George, then others like Sting and Bono.

Money stuff happened, PY became a waiter in a restaurant he once owned.
Now known for ages as the Tex Mex band Pacaminos guy.

[Sam Spoons]

There were two Brit singers called Paul Young, I think you're referring to the younger (still living) one who had a hit with "Wherever I Lay My Hat". The other was lead singer with Sad Cafe and co-lead with Mike And The Mechanics (sharing with Paul Carrack).

[alexis]

https://lalals.com/voice/rod-stewart

Haha. I wish I could audition a free sample now without having to go through the trouble of uploading a vocal of my own.

I'm just getting my system up and running so, though I have an idea of where I could use AI Rod's help, I don't have a vocal to send him yet.

Imagine if AI Rod sounded good, and that Vocal morphing tool (that Drew referenced at the top of this thread) worked well to blend my voice with AI Rod.

That would be the Holy Hand Grenade, for sure.

[amanise]

There were two Brit singers called Paul Young, I think you're referring to the younger (still living) one who had a hit with "Wherever I Lay My Hat". The other was lead singer with Sad Cafe and co-lead with Mike And The Mechanics (sharing with Paul Carrack).

[adrian_k]

See, I am of a different point of view (I often am ), but your voice is your voice and it’s kind of part of your identity, so I wouldn’t want to mess with it. I guess it depends why you make music.

[Matt Houghton]

https://lalals.com/voice/rod-stewart

Haha. I wish I could audition a free sample now without having to go through the trouble of uploading a vocal of my own.

I'm just getting my system up and running so, though I have an idea of where I could use AI Rod's help, I don't have a vocal to send him yet.

Imagine if AI Rod sounded good, and that Vocal morphing tool (that Drew referenced at the top of this thread) worked well to blend my voice with AI Rod.

That would be the Holy Hand Grenade, for sure.

I tried lalals a while back when I wanted to lay down some bvox in a different voice from (my own) lead, just to sketch out some ideas. worked pretty well tucked into a mix. But I found that unless the voice is a pretty similar range/style to your own, you have to adjust your own singing to get the best results from it — no good just recording a part singing exactly as you always do. I didn't try to emulate any actual person's voice, mind; that holds zero interest for me!

[alexis]

See, I am of a different point of view (I often am ), but your voice is your voice and it’s kind of part of your identity, so I wouldn’t want to mess with it. I guess it depends why you make music.

Points well taken, thank you, adrian_k! On the other hand there's a long rich history of changing voices, and in a myriad of ways (EQ, doubling, chorus effect, pitch shifting, adjusting formants, distortion, running it through a Leslie or a guitar amp, etc.).

I understand and agree to a degree with someone (not necessarily you ) feeling that AI voices are in a completely different category of music making, but I also feel that in a sense they are just another toy/ tool to play with.

Similarly, I understand someone feeling that there's something mildly fraudulent about using AI voices or instruments, but IMO that horse left the barn a while ago, when "world class drummers", and keyboardists and bass players started living in our computers, with apps like EZdrummer, EZkeys. EZbass, etc.

(Or maybe that barn door swung open when we started using samples of instruments we don't actually know how to play ... ?)

[Wonks]

Just wait for EZvocalist to come out then. I’m sure it can’t be long!

[alexis]

https://lalals.com/voice/rod-stewart

Haha. I wish I could audition a free sample now without having to go through the trouble of uploading a vocal of my own.

I'm just getting my system up and running so, though I have an idea of where I could use AI Rod's help, I don't have a vocal to send him yet.

Imagine if AI Rod sounded good, and that Vocal morphing tool (that Drew referenced at the top of this thread) worked well to blend my voice with AI Rod.

That would be the Holy Hand Grenade, for sure.

I tried lalals a while back when I wanted to lay down some bvox in a different voice from (my own) lead, just to sketch out some ideas. worked pretty well tucked into a mix. But I found that unless the voice is a pretty similar range/style to your own, you have to adjust your own singing to get the best results from it — no good just recording a part singing exactly as you always do. I didn't try to emulate any actual person's voice, mind; that holds zero interest for me!

Interesting history, thank you for sharing that, Matt!

Re: the part in bold ... are you saying it might be good to sort of try to channel the AI singer of choice and meet them half way as you lay down the guide vocals?

[alexis]

... I guess it depends why you make music.

Though I love to hear my voice on tape (and hate it at the same time!), one big reason I'm interested in all this is to get a "real" singer to do songs I write, as a cheaper alternative to hiring live singers.

In the past, I hired an online human female singer to sing a duet "with me", I was very happy with those results. (On the other hand, I was disappointed with results of an online human singer doing one or two other songs I'd written, oh well).

Lately I've been sitting in with a band for a few months and have started exploring asking one of the band members to do just that as well. Very excited about that possibility! (One unintended, and painful yet welcome consequence of this is the realization, or at least the fear, that most of the songs I've written may not be good enough for someone else to want to sing. Time to up my game! ).

[Matt Houghton]

you have to adjust your own singing to get the best results from it — no good just recording a part singing exactly as you always do.

Re: the part in bold ... are you saying it might be good to sort of try to channel the AI singer of choice and meet them half way as you lay down the guide vocals?

Kinda.

I don't know if/how lalals has evolved since I tried it. But more about making sure you're in the right range for the singer in question, how much your voice is in your chest v your head, and how they sing vowels in particular (eg. is it a short 'a' in 'dance', as one would use across the north of England, or a long 'a' that sounds like 'dahn-ce'. And if you want to go to extremes and have a male vocal replaced by a female one, you might have to work harder at the pitch and chest v head stuff...

I'd stress that I haven't used this a lot, so take this all with the pinch of salt it deserves — and maybe try it out and post back here with your own observations; I'd be interested. But I did reach these conclusions after a day or so's experimentation. I was basically looking to get a feel for how useful it could be for fleshing out demos of new songs. And I reckon it could be quite useful, or possibly even very useful if you just want to lay down things like basic oohs and aahs in different voices...

[amanise]

I think one thing you can be sure of is that when the inevitable happens and 'Easy Vox' does come out (not to want to appear to diss any existing trade names) - it's going to be expensive. Everything else about a commercial track seems to be a bit negotiable with the listener - but the main vocal? Nope. And for weird reasons too! Sometimes very far from musical - Lou Reed anyone? His mixing engineers needed a medal of honor each. The guy had a stellar career though.