A lot of UTAU resamplers make vocals sound sharper, stronger and more crisp than the actual recordings. SynthV AI only recreates the voice as it is, while neutralising the way vowels are carried due to its model. So, Xia Yuyao having a softer sound isn't surprising, but now I'm worried she's going to be nothing but soft.
All but one of her vocal modes are focused on producing gentle vocals. There is a chance Solid could carry her stronger tone really well, and we already had a soft vocal who surprised us with her potential power before - Chifuyu.
I will hold onto a last bit of hope that Xia Yuyao will be more like Chifuyu and less like ANRI. That is to say, more versatile.
All but one of her vocal modes are focused on producing gentle vocals. There is a chance Solid could carry her stronger tone really well, and we already had a soft vocal who surprised us with her potential power before - Chifuyu.
I will hold onto a last bit of hope that Xia Yuyao will be more like Chifuyu and less like ANRI. That is to say, more versatile.
I got my box and shipped the other copies to my friends (they'd sent me money and we'd ordered 3 since I had an account or something, I forget the exact reason). She's lovely! Comes with the code already, so we're ready for when the actual bank drops. I can't wait for proper demos, but I'm sure I'll love her either way. Especially interested in that Dark vocal mode.
we just need light and vivid
I remember them toying with the idea that they might not keep the UTAU vbs around, I checked her page on their website and the whole section that hosts links to her utau vbs and image downloads was taken down, however, I searched in my history and the vbs themselves are still up, just in case go ahead and download them:
Bundle containing all her CHN voicebanks
Her JP vb
Bundle containing all her CHN voicebanks
Her JP vb
Idk she’s kinda REALLY boring sounding:(
this doesnt showcase the sweet and dark modes well i think ... Like they imo dont sound that much different from her default mode. I like her Solid and Whisper though!
I heavily disagree regarding Sweet. I think the difference is fairly audible.
Sweet and Solid are my favorite, I think they sound most like her UTAU voicebank. Overall, very pleased with her vocal; I think she sounds great and will be very versatile
Mhhhh, I don't know how to feel about her. I was honestly expecting her to sound way different (like, I completely understand UTAU and AI vbs being distant under every aspect, but she sounds too different to me; if I close my eyes I can hardly think I'm listening to Yu Yao sjhx). She doesn't sound bad, at all, she just doesn't sound like Yu Yao to me.
That’s about where I am. She doesn’t sound bad! I’m just… personally I’m unimpressed. I kinda have this feeling of “…that’s it?”
I'm not familiar with her UTAU VBs, but listening to her SynthV VBs, I don't think she stands out a whole lot among the other Chinese voicebanks, barring perhaps Solid.
Well, majority of her UTAU voicebank's tone was due to the program itself. Listening to her raw recordings and comparing them to the AI vb (especially solid or sweet), you can definitely hear the voice provider's natural sound correlation.
This was always going to be an issue; sample-based voicebanks just don't carry over to AI well, especially when the programs are completely different [Vocaloid>SV, UTAU>SV, rather than SV(std)>SV(AI)]*
*this is not to say, however, that attempting to maintain similarity is futile, it most certainly CAN be done
--
I don't think she's necessarily innovative, but I would definitely place her behind Stardust and in front of the DT vocals. She has a really good amount of vocal modes, a cute character, and much more engaging all around than the alphabet-series in my personal opinion.
This was always going to be an issue; sample-based voicebanks just don't carry over to AI well, especially when the programs are completely different [Vocaloid>SV, UTAU>SV, rather than SV(std)>SV(AI)]*
*this is not to say, however, that attempting to maintain similarity is futile, it most certainly CAN be done
--
I don't think she's necessarily innovative, but I would definitely place her behind Stardust and in front of the DT vocals. She has a really good amount of vocal modes, a cute character, and much more engaging all around than the alphabet-series in my personal opinion.
Agreed with this whole sentiment. Xia Yuyao is already more notable than any Dreamtonics voicebank just by having a design. Simple demos like that show what the basic voice sounds like, but are terrible for judging how interesting the voice is. I wonder what her crosslang will sound like.
Her voice is nice, it doesn't stand out a whole lot from DT's other voices (very much normal girl with a nice voice singing normally), but I could see people doing great things with her. I wish her dark/sweet were more noticeably dark and sweet, respectively.
I definitely prefer her SynthV to her UTAU, mainly due to the reduced engine noise, which I find super bright and grating in her original vb. Appreciate the accapella demo for its clarity, hoping to hear how she sounds in full songs soon! I only pick up new voices if they do something my current lineup can't, so I probably won't be buying her; but I'm glad a classic character gets to live on.
I definitely prefer her SynthV to her UTAU, mainly due to the reduced engine noise, which I find super bright and grating in her original vb. Appreciate the accapella demo for its clarity, hoping to hear how she sounds in full songs soon! I only pick up new voices if they do something my current lineup can't, so I probably won't be buying her; but I'm glad a classic character gets to live on.
Is there anything special about her voice from a linguistic perspective? I know she's a Taiwanese VB, so are there any characteristics of her voice specific to her country of origin? (From what I know Mandarin is the official language of Taiwan, but I imagine there are accent distinctions?)
According to my friend she has Southern Chinese pronunciation! He likes it because it's close to his accent.
Well, Taiwanese Mandarin and Mandarin from mainland China are fairly different. For example, retroflexes are typically changed to postalveolar consonants. Another example is that [f] may become more similar to [h] and the nasal endings of -n and -ng might merge, making [zheng] and [zhen] into homophones.
To make it even more complex, Taiwan (legally) has 3 national languages: Mandarin, Hokkien, and Hakka. Hokkien and Mandarin are the most prevalent of the two, but all three are required on public transport signs, etc. Some families speak Hokkien more at home, some speak Taiwanese Mandarin more at home, and these differences can cause some bleeding across the languages for different accent variations.
Ooh I’m super excited for her!! Xin Hua remains my favorite Mandarin voicebank and having another Taiwanese accent option is super exciting! I also love her whisper vocal mode and the idea of her dark one… I just hope it’s actually you know different than her standard vocals