Ahh!!! OK great thanks!
Not seeing any change in behaviour here. What situation are you running into this up/down pitch problem? Can you provide a screenshot?
If you get frustrated by a beta, maybe using a beta is not for you. The whole point is to iron out problems before release.
1 Like
Korean Singer Uni is not working properly. Please check it.
Ok cool. Thank you for sharing
Yep 
. Someone said it is hidden behind ui
1 Like
Apologies for the late reply, but I’m kinda flip-flopping between their 201s and 205s to try and figure out if it’s just me being a little picky, or if 205 genuinely isn’t hitting the same for me. I keep second guessing myself a lot, because maybe I just wasn’t used to their new versions and it’s not too bad. I dunno if we can send videos here in replies or not, but later I might just so I can get a second opinion. Still can’t tell if I’m just being picky or whatever lol.
I definitely get what you mean. There is some subtle changes.
I am a Korean speaker, and I am preparing to send Dreamtonics some feedback and bug reports regarding this beta update in the near future. Since there may be other Korean speakers like me, or people who make songs in Korean, I would like to share part of the report I intend to send to Dreamtonics.
- Issues with the implementation of certain Korean phonemes due to changes in the phoneme system
In Korean, in addition to the phoneme “s”, a new phoneme, “s\”, has been added in this beta update to represent the consonant ㅅ.
I am not very good at describing the nuances of my own language, so this is the best way I can put it, but I am satisfied with the sound it produces, as it comes closer to the way native Korean speakers normally pronounce the consonant ㅅ.
I also found it quite satisfying when I selected the v205 b1 voice and listened to it. It seems to be represented as “s\” in X-SAMPA, or as “sy” otherwise.
This is where the problem occurs.
In the V201 voices, it seems that the phoneme “sy” is recognized as the X-SAMPA phoneme “s”. When I enter the Korean syllable “셔” (syeo), the phonemes “sy eo” are entered, and the resulting pronunciation is “서” (seo).
To work around this, I have to insert a space between “s” and “y”, entering them as “s y” so that they are treated as separate phonemes. This resolves the issue.
Based on this behavior, I suspect that the Korean phoneme “sy” (“s\”) is being incorrectly mapped to “s” in the V201 voices, which would cause it to produce the pronunciation associated with the “s” phoneme.
However, I cannot determine whether this is actually the case, because the new X-SAMPA option is disabled for the V201 voices in this beta version, meaning that I cannot enable it for those voices. As a result, I cannot tell whether the phoneme “s\” is being applied correctly but rendered incorrectly, or whether it is being incorrectly substituted with another phoneme, as I suspect.
1 Like
I don’t know about the Korean phonetics but I’m curious as to where you got information about the phone and the structure for them
Correction: phomemes not phone
i mean Samsung IS a Korean brand. So if you do actually need to find information about their phones, they have a web site that will help you 
Sorry. I just had to do this. 
I’m always messing up my spelling or letting autosuggestions take the lead, which ends up me using the wrong word
Information on the phonemes and their structure up to previous versions can be found on the official Dreamtonics website. As for the new phoneme system introduced in the current beta version, if you check the Dreamtonics 2.3.0 b1 beta update announcement, you’ll find hyperlinks to documents that describe the new phonemes and their structure.
Thanks for the info.
V201 voices do not support the new phonemes, such as [s\] for Korean starting from V205b1. Third-party voices are pending an update for the final release of SynthV 2.3.0.
You’re right. My guess is that, in 2.3.0 b1, certain Korean pronunciations may be rendered differently in the beta editor when using V201 voices because phonemes that V201 doesn’t support are being entered by default, regardless of the voice version. The issue itself can be easily fixed by manually correcting the phonemes.
I found a bug. The Korean consonants ㄲ and ㅋ are allotted the wrong phonemes, namely kh and k_t respectively. They should go the other way round; k_t for ㄲ and kh for ㅋ. Likewise for ㄸ/ㅌ and ㅃ/ㅍ. ㅉ and ㅊ are correct though.
Also, the phoneme allotting for the coda ㄲ also is weird. Examples are “섞어” being s V k g V and “볶아” being p o k g 6. They should be s V k_t V and p o k_t 6 instead.
EDIT: I’ve sent this as a bug report to Dreamtonics.
1 Like
They say they have already noticed this issue and are fixing it.
at least we now have some of the German phonemes so it’s easier to work out what’s going on. I find German useable in its current state but certainly no where near native speaker standard, In fact I possibly got better results using Spanish as the model earlier but with far more effort – at least now I can get something useable with very little input and it’s not really worth putting in the work at this stage as no doubt things will improve.
UPDATE just installed the second beta and certain sounds like “r” are noticeably improved and “u” is more reliable “ö” still sounds like “o”, though but if I change to “oe” instead of “OE” then that’s also much better.
1 Like
I can confirm that the German pronunciation has improved in the current beta versions of the Dreamtonics voices.
Since I use these voices primarily to create practice mp3 demos for human singers, I can live with the fact that they still retain a foreign accent.
To improve the pronunciation, I am constantly building and refining my own German-language dictionary. Most of the entries in my dictionary begin with a vowel, and I add a “cl” prefix to these to avoid the often unnatural-sounding transition from the preceding syllable.
While this can also be achieved by placing an apostrophe at the beginning of the word, given that this is a recurring issue, it would be beneficial if the German dictionary provided by Dreamtonics were updated accordingly.
Have you figured out a way to make a ch as in “ich” to sound remotely like a native speaker? Would love to know, if you have… Thanks!