I should probably work on the harmony vocals a bit more. I was trying to keep them… well, background vocals. But a bit more expression at the end of phrases might be needed.
And maybe a bit of de-essing. It’s a bit of a balance between clarity excessive sibilance.
Edit: bellerandre pointed out that I should used the SynthV feature of splitting out the harmonic and inharmonic parts into different channels or files. So I’ve done that and re-uploaded the file.
Thanks, done! I knew about the feature, but was too lazy to actually try it out.
I used the separate aspiration for both the lead and harmony, and used Spitfish to perform the de-essing.
The main problem was that I was using a really aggressive compressor (Vulfpeck), which sounds great, but can do nasty things to sibilance. Having it not hit the compressor was probably enough, but I used a de-esser just in case.
It’s also good to pull it off the background vocal, as it keeps it out of the reverb.
However, the transposed voice gives me a distinct ‘uncanny valley’ impression – is it a gal or a guy? Sometimes the one, sometimes the other…
At 2’37, the word “happli” (‘happily’) is slurred over, beacause the lyrics writer has crammed it into too few beats – obviously an important word that ought to have got a sufficient number of beats to be sung ‘happ-ee-lee’. All IMHO, of course.
However, the transposed voice gives me a distinct ‘uncanny valley’ impression – is it a gal or a guy? Sometimes the one, sometimes the other…
Well, when and English male voice is available for SynthV, I’ll be first in line.
But, yeah. One of the problems with leaning into the masculine territory is that it get mush-mouth at a point when the formant drop too low.
SynthV does a really good job altering the source vocal, but you can on tweak it so far before you get some sort of artifacts.
At 2’37, the word “happli” (‘happily’) is slurred over, beacause the lyrics writer has crammed it into too few beats – obviously an important word that ought to have got a sufficient number of beats to be sung ‘happ-ee-lee’. All IMHO, of course.
Noted. I considered that, but thought I’d go with the more conversational pronunciation, which is how I normally mispronounce the word.
Plus, adding the other syllable made the line sound more fussy and cramped than it already way.
But… if I may have another look at it at some point.
I appreciate you taking the time to listen and comment, thanks!
Very interesting work! Eleanor sounds pretty good as a guy as well (wow?!)
Well, since the underlying mechanisms are basically the same. it should mostly work. There’s a very minor difference in the vocal cords, and the primary difference in the vocal tract is the length. That results in a shift in pitch and position of the formants, which is what the gender parameter does. But shifting too far past the original values obviously degrades the voice quite a bit.
There’s currently no way within SynthV to emulate the glottal effect of roughness, although there was in the prior version. But the interface to that was probably too technical, requiring too much knowledge of speech production. I don’t know how may speech pathologists are using SynthV.
But I hope that feature makes it back into SynthV with some usability considerations. It would be really useful for a lot of pop voices, and I think the output of SynthV is likely to be much better than Voclaloid.
In the second half of the song, his/her singing sounds smoother to me but as a strange consequence, also sounds less ‘guyish’. Could just be me.
Interesting observation! I’ll have to listen for that.
– obviously an important word that ought to have got a sufficient number of beats to be sung ‘happ-ee-lee’. All IMHO, of course.
always nice to see some creativity!
