Is there a way for me to correct this once globally or do I have to edit every instance?
If you know the alternate phoneme sequence you want to use, create a dictionary and add the words you want changed:
Remember to then select that dictionary on any tracks using Hayden.
Also keep in mind that “the” is contextual based on whether it precedes a consonant or a vowel. Overriding it to correct the “thuh” pronunciation that goes before consonants will also change the phoneme sequence for the “thee” pronunciation that preceeds vowel sounds, so you may have to literally type “thee” for those lyrics to sound correct.
「いいね！」 2
I’m using Hayden version 101b1, and not hearing this.
“The” converts to dh ax, and “these” to dh iy z.
Perhaps there’s something else going on? 
Perhaps Hayden is visiting New Jersey? 
I have had to separate the preceding word or add a glottal ‘cl’ to get a clear ‘t’. And THEN, close it back up so the AI fits them together. Once it’s there, don’t touch it.
But I am still very much learning how to deal with the t,b,p clarity, and just test till it breaks or works.
「いいね！」 1
@The_Soundsmith Could you give me the steps to try this? I’m not clear on if you mean two words in one entry or… I dont understand. Now I am trying to have Hayden sing “roads” and it’s “rose” every time. The only way I can get the “d” sound is when I make it “road”. As soon as I add the “s”, it audibly becomes rose. Frustrating.
@dcuny I don’t know. Try “winding” “roads”. For me, it clearly says rose. It says rowdz above it. However, audibly, it 100% sounds like “rose”. I don’t know much about phonemes and I imagine if i can’t get it right on the timeline, then I don’t have a correction for the dictionary anyway.
When that happens to me, I try the Note Properties tab and play with the syllable Duration and Strength. I try to keep the total Duration at 100% or less, some words will lose the last part of the sound at more. ‘rowdz’ make the ‘d’ longer, reduce the time of the ‘o’ and give it to the d. Then make the d louder (140% or more might be needed.) That’s a beginning. Play with it and learn, it’s all i can do and eventually I find things that work. If you use the word a lot, put it in the dictionary. Good luck.
Here’s an example with Hayden:
If you look at the waveform for roads, you’ll see there’s a reduction in the detected pitch in the roads before the final /s/. That’s where the tongue taps the top of the mouth and to create a short closure or “stop” of the d.
Look at where the pitch line drops, because there’s no voice/pitch at that point. That’s the “stop” before the /d/ sound.
It’s not that strong, but it is there. If you change the duration of the /d/, you’ll get a longer and more clear stop:
There’s a drop in the pitch line in rose, but that’s at the /s/, which is unvoiced (is has no harmonic content).
「いいね！」 1
Thank you all. I was experimenting with exactly this last night. I did not get it nice and clear but I did get more towards “roads” instead of just “rose”. It’s more of a distorted “t” or something but at least a consonant is in there and a listener would more likely interpret a “d” in context with the words. I look forward to learning more and / or hopefully this will improve soon. @dcuny , are you on the beta team or does anyone have access to the betas?
It looks like there needs to be enough time to articulate the phoneme.
At 120bmp, if I do this, the /d/ won’t get articulated properly, and I get “rose”:
Changing the Strength of the /d/ to 180% makes it noticeably better.
Even shortening the first syllable down to a 16th note, the /d/ is still clear:
「いいね！」 1
I have 1.10.1 and Hayden 100. No updates available when I click “Check for Updates”.
Yes, 120bpm is exactly what I am creating with. I will keep experimenting with your examples and I appreciate your useful information.
Beta versions cannot be installed from within the editor because they are likely to have bugs or instabilities.
Check the latest news post, links are at the bottom: