Edit: I’ve changed the key in the first half of the song so it’s not so far out of Eleanor’s range.
Both the female and “male” voices have been doubled with the non-AI version. This was mostly to add some clarity to the “male” AI voice, as it’s a bit unfocused in the lower register.
There are some places where the vocals could be a bit more clear, so I’ll probably go back and work on that. Fixed! But feel free to tell me if you disagree.
I’m obviously going for a '70s sort of low-fi vibe here, so apologies in advance for the “crunchy” mix.
Note: I hear I didn’t pan the female aspiration when I panned the voice. Gotta fix that, too! Fixed!
And I can’t wait for a “real” male English voicebank becomes available!
Thanks, I appreciate that. It seems that morphing female vocals into male vocals works than the other way around. It’s certainly a usable voice, but there are a few vowels that suffer in the process.
The voice isn’t as “clear”, in the sense that it doesn’t have the same harmonic resolution as the standard voice. Have a look at it in a spectrogram, and you’ll see what I mean. It looks fuzzier, and it sounds fuzzier. It also seems to have less resolution in the time domain as well.
It also seems to respond to parameters differently, although I haven’t quantified that in any way.
But the voice is certainly very appealing, and I’m happy to have it.
I don’t know if Dreamtonics is planning on discontinuing the classic voice approach or not, but the ability to choose alternate takes is very nice, and I’m still not entirely clear what the “strength” knob does. Does a “less strong” vowel sound more like a schwa?
Some more documentation - that is, more than zero documentation would be helpful.
In any event, I’ve gone back and made some small changes to some of the lyrics so (I think) they are a bit more understandable, and boosted the guitar solo so the song sounds like it’s got a proper intro now.
