As I finally bit the bullet and decided to support Cantai, despite certain misgivings at this stage, I managed to find a project it actually renders and I think this one might actually be among the more suitable for Cantai with its current voices. There’s quite a lot to dip into here as Songs 2, 3 and 5 are set in full from the Biblical Song of Songs with a string accompaniment at times. Although it’s really designed for a small choir, the Cantai choirs are currently too crude to be useable in this context so I chose Soprano=Hannah, Alto=Violetta, Tenor=Rigoletto and Bass=Hidenori.
With the notable exception of “beloved” – something which I think has already been mentioned and is of course understandable, I needed to change very little in the text pronunciation. I’ll leave it open for others to comment on any other issues or strengths they might find and of course with luck, one or two might enjoy some of the music as well.
I realised that didn’t use the NPPE version of the CSS strings which was a mistake as it sounds much better (and was used in my SynthV version). So the link above has now been updated with the new version - with a few further corrections.
Wow, I’m impressed. Would you have played this to me without mentioning anything, I would have believed this to be real singers / players. Truly amazing.
I know, I know, the expert listeners will have a lot to remark and complain, still, I’ve been following the Cantai discussions here only a little, and I did not expect the quality of Cantai to be that high. I’m blown away.
If someone had said to me just a few days that it was possible to get this with Cantai, I wouldn’t have believed it either. After all, it’s only yesterday I actually decided to invest in it having been hopeful but sceptical up to now. I still think the choirs are of little use (plans are afoot for new ones) but in this particular music where a little passion and volatility is required, I actually prefer it to my existing SynthV rendering and it has now replaced it on my Reelcrafter site (where I will put any further updates and refinements).
Text rendering is in general very good – it doesn’t sound too American. There are still some issues with dynamics – the individual singers are indeed very individual and require different dynamic levels (and even ideally different Expression maps purely to set the dynamic range as the EM’s otherwise do nothing). On occasion, they don’t respond as expected but things in general have moved on a lot since the first beta – now it would be nice if I could render more of my existing projects but I’m sure that will be fixed in time. When it works, it can do so rather nicely, given the right sort of piece!
Nice work. The only thing that really caught me by surprise was the immediate presence of the alto voice later in the piece as if the singer were literally in my face.
It may just be me but it seems to begin around 18:32 or so into the recording. It’s not awful by any means yet it’s as if the alto came to the edge of the stage after being about half way back previously.
It reminds me of a singer I once accompanied during undergrad days who would walk around the piano and manage to drop her head next to mine at the loudest part of the piece just to annoy me. We had one of those love-hate relationships between singer and accompanist that was always full with what we called the “poor choices.” In kind, though, I often started our rehearsals playing a half step or whole step above the written key which was misery for her because she had perfect pitch - poor choices.
although the alto is supposed to barge in a little to the male meanderings, you’re quite right – it’s much too blatant and I’ve corrected this now. As well as making a little change to the overall stage positioning in Dorico which might also help a little in general.