I have a little issue. Synth V is great but on this one particular line it will not, no matter what I seem to do pronounce the T in the word Twist after “Like the”. It omits the T and says “wist”. If I remove the “Like the” phrase then it pronounces the T. I’ve tried uncheck relaxed consonants and it still does the same. I’ve also tried different takes. Any ideas? I’m quite new to this and am just sketching out some vocals but pronouncing the T in Twist is pretty essential.
I have just created the phrase “like the twist” using Solaria lite and then again with Solaria full version.
If there is a tiny “sil” between “the” and “twist” then she does indeed say “wist”, pronunciation slightly different for each version but both failing to pronounce the leading ‘t’.
Without any gap (no “sil” visible) both voice banks correctly pronounce the sentence.
Could that be the problem in your case?
I also increased the silent gap between "the and “twist” and pronunciation improves with length of silence. I guess you may have an unintentional small silence or one too short for the rendering to cope?
Thankyou for the reply. No there is no silence between the words. The only thing it does do is the word “the” is 2 semitones lower then goes back to the original note for the Word “Twist”
Could be a speed thing. The BPM is only 90 and the singing is no quicker than you’d naturally say it. I’ve tried increasing the length of the "the and the “like” and it still does the same lol. It’s driving me mad lol
I just tried with tempo at 90 and 110, my results are broadly the same in both cases:-
A small silence causes the problem…
a larger space gives good pronunciation…
no space give good pronunciation…
I believe that the rendering engine is trying to pronounce the ‘t’ of ‘twist’ before the note onset, clearly visible in the bottom image, it looks to be half-way through ‘the’, when there is a small silence the rendering engine finishes pronouncing ‘the’ and only then tries to cram this leading ‘t’ in before the note onset - a gap too small so it fails. If the gap is big enough then it can again be pronounced prior to note onset (second image).
Thanks again. The only way I can explain the issue is by showing it I guess. My “Like the” are quite quick but no quicker than you speak them or say the phrase “Like the twist”. So here is a video to exactly my issue. This is a default new Synth V file. Not changed a thing.
Be interested to know yours, or anyone elses thoughts on this one as it’s bugging me now lol
Exactly the same for my examples.
What I am hearing in your video sounds just like I had for my screenshots and you do have a very short duration silence between those troublesome words, that gap is too short for the engine to cram a leading ‘t’ in.
So, I think you either need to consider extending the duration of ‘the’ such that it bumps into ‘twist’ (it will start to ‘push’ the note for ‘twist’ out of the way) or …
… and this is a long way from perfect…
… spell ‘the’ as ‘thet’, it will then pronounce that trailing ‘t’ almost as if it were the leading letter of ‘twist’ — it might work in your case - if the gap between words isn’t long enough to give a discernable silence ???
…there is no gap between them at all between the “the” and the “twist”. As you can see from the video. If there was there would be an SIL notification between the two and you can see in the video there is not. And it snaps to the grid anyway especially at a low resolution grid. So that can’t be the issue!
Secondly extending the “the” wont fix it as I’ve tried and obviously the vocals won’t scan correctly if I do that. Extending it even a little throws the scan clean out and it STILL won’t pronounce the T lol
I’ll try you’re suggestion of writing “Thet” instead but I feel that may not work lol
It’s a shame as this little issue has kinda rendered the synth pretty useless for me as I feel this is not a difficult line for the synth to sing and everything else seems fine. Maybe it’s a bug???
Anyway I totally appreciate your help and time. Thank you. If you or anyone has any other ideas please feel free to let me know. As I really do like this Synth!
Well, the only other thing I can suggest is select ‘show phonemes’ so you can see what is being pronounced where, as below
You can then clearly see where the probematic ‘t’ is being placed on the time line (and yes, that is how we would naturally sing such a phrase), maybe you can figure things out a bit with that additional detail.
Ah ok, here I think is the issue. I’m pretty sure it shouldn’t do this so maybe it is a little bug. As you can see in the picture there is a huge dip in pitch right over the T of twist. If you manually draw the line back in the T re-appears. Although not perfect but certainly usable
Your pitch curves are pretty violent there - compared to my screenshots showing default pitch curves - no tweaking by me - which sounded lifeless - vocoder-like. One extreme to the other!
I started from scratch for this three-word project so cannot quite see what could cause this difference but I guess that’s a side issue, the main problem is now solved. or at least a workable solution found.
Yeah it’s pretty crazy isn’t it. Didn’t do anything other than type the words in so the program put those curves in itself. Very strange. Either way a solution has been found. Thanks for all the help
This is interesting, I haven’t encountered this before myself but I see the issue when I put in that same phrase. It also seems to happen with all the voice banks I have, not just Solaria.
To me it sounds like the closer the two words “the” and “twist” get, the more relaxed the “t” consonant becomes. However, unticking “Use relaxed consonants” didn’t change the situation at all.
It seems to help greatly if you increase the phoneme Strength and Duration sliders for the first “t” of “twist” to 180% and reduce the “ax” at the end of “the” to about 80%. Still not quite the same as the word on it’s own, but much improved.
It certainly is an intriguing one. I do feel it shouldn’t really struggle with it. It’s not like it’s singing really fast or anything is it but it is dropping the pitch crazily. Drawing in the pitch will bring the T back but I feel the crazy pitch drop is minor bug
Yes I agree, it’s better but not perfect. I tried another method by dividing the word “The” into two notes to add a “-” to listen. See the picture below:
Yeah that certainly makes it better doesn’t it. Good idea. I still feel like manually drawing the dip of the curve back to the line gives the best result though but this is certainly much closer without editing the curve
or at least a workable solution found.
