Trying an AI voice in an old track.

[zerocrossing]

A bit of background. I wrote this song with a friend in the 80s. I later went back to it to "finish" it, but my brain wants melodies that my vocal chords won't do... and frankly, my vocal chords don't really do any melodies that well. The only time I've ever had this song the way I intended it to be was when I had a vocalist to collaborate with, which I'm still open to, and prefer, but I long ago gave up on that.

So, Synthesizer V. I bought it a while ago and never used it, but ReSing made me think about it. It's pretty tedious to use, but the results are a lot better than ReSing, and because I can put in the exact melody that I want, it's a lot better for my needs. Everything else is performed by me.

[wagtunes]

I know a lot of people will avoid this thread but I couldn't wait to hear this. But isn't Synthesizer V considered more synth vocals than AI vocals? I do admit, they sound good. But when I think of AI vocals I think singing the song yourself and having your voice replaced with an AI voice through processing.

Anyway, whatever you want to consider this, it sounds great.

[eassae]

I tried Synthesizer V when it came out and was impressed with the results. It's like having a good but not genius singer in your daw. It's my understanding that all the voices start with a real singer, but there is some AI thing going on in the background to make the output more expressive. There's ads on kvr right now saying they're adding basically the same tech to the EastWest sound libraries. I was curious to see how the vocals I created did in AI testers and the tracks I did came back as 89% likely that they were human made which is interesting. I agree with the OP that it's a bit of a beast to work with. I spent hours tweaking the phonemes and pitch fluctuations, but again good results. I also did a bunch of layering and processing to the original output. Ultimately, I did only experiment with a few tracks and I didn't upgrade to v2, because when I heard the demos they actually sounded worse than v1, having more of the artifacts associated with AI vocals. The Synthesizer V demos ironically really aren't the best examples of what the software can do though.

If I didn't know beforehand that this track used synthesizer v, I don't think I would have suspected non human vocals. There's a few spots if you listen very carefully that you could possibly tell 2:45ish being one, but it's not clear that it's just normal distortion. It's a cool track.

[The King of the Sofa]

Funny

[zerocrossing]

I know a lot of people will avoid this thread but I couldn't wait to hear this. But isn't Synthesizer V considered more synth vocals than AI vocals? I do admit, they sound good. But when I think of AI vocals I think singing the song yourself and having your voice replaced with an AI voice through processing.

Anyway, whatever you want to consider this, it sounds great.

I think they do create a voice model that’s based on neural networks. I might be wrong. But thanks!

[zerocrossing]

There's a few spots if you listen very carefully that you could possibly tell 2:45ish being one, but it's not clear that it's just normal distortion. It's a cool track.

Could be pegging the UADX 610 preamp emulation I have it going though... not sure. I didn't really notice it.

[eassae]

There's a few spots if you listen very carefully that you could possibly tell 2:45ish being one, but it's not clear that it's just normal distortion. It's a cool track.

Could be pegging the UADX 610 preamp emulation I have it going though... not sure. I didn't really notice it.

It's only because I tried to listen super close to see if I could find any anomalies. But yeah, definitely sounded like it could be an some sort of coloration not artifacting.

[zerocrossing]

There's a few spots if you listen very carefully that you could possibly tell 2:45ish being one, but it's not clear that it's just normal distortion. It's a cool track.

Could be pegging the UADX 610 preamp emulation I have it going though... not sure. I didn't really notice it.

It's only because I tried to listen super close to see if I could find any anomalies. But yeah, definitely sounded like it could be an some sort of coloration not artifacting.

There is a point where I think I forgot to change a note in a harmony and you get two identical words that are phasing. At first I was all, "gotta get that fixed," but it sort of grew on me, like an effect I put on it.

[eassae]

There's a few spots if you listen very carefully that you could possibly tell 2:45ish being one, but it's not clear that it's just normal distortion. It's a cool track.

Could be pegging the UADX 610 preamp emulation I have it going though... not sure. I didn't really notice it.

It's only because I tried to listen super close to see if I could find any anomalies. But yeah, definitely sounded like it could be an some sort of coloration not artifacting.

There is a point where I think I forgot to change a note in a harmony and you get two identical words that are phasing. At first I was all, "gotta get that fixed," but it sort of grew on me, like an effect I put on it.

It's not something that bothered me. I listened again an it sounds intentional.

[ChameleonMusic]

Interesting, full on energising track! Enjoyed it.

Don't have an issue with someone writing / recording their own music and then using such AI vocal models...not everyone has a friend who can sing professionally / afford to hire a vocalist + studio etc.

VERY different from someone typing a few words into an AI music website and ''creating'' the whole f*ckin' thing from scratch and then claiming they're a composer! This here is not much different to using a vocal sample library, really.

Vocal line works well and is highly appropriate for the song, but, for me, sounds AI at many moments throughout the performance...not that's an issue any more than Tina Turner sounding like TT or Jon Bon Jovi sounding like JBJ!

Played it to a colleague (very experienced music tech and music-wise) - first thing he said was ''is the voice AI?''
It just doesn't sound quite 'real' at times - pitch too perfect on occasional sustained notes...ditto the timing...mini pauses at certain points that would be there with a real singer are missing...pitch changes / slides / glides are all a bit too smooth...pronunciation of certain words is a little too clear etc...

Having said that, I suspect 95% of people listening for relaxation / pleasure would not notice at all + it doesn't really matter very much as it works! Definitely better than having me sing it!