Unpopular Opinions

[lIlI]

Is it possible that Yamaha might still be using concatenative synthesis, with Vocalochanger being the only AI element? Or perhaps they're training the AI off concatenative voicebanks, to keep that classic Vocaloid sound - similar to people who create 'AI' versions of concatenative voicebanks by running them through Diff-SVC. I mention this because, by default, the V6s still sound very much like concatenative voicebanks. Fuiro sounds exactly how I would expect Philo's voice to translate to concatenative synthesis, at least from what we've heard so far, and the English V6s have the same pronunciation quirks as previous English Vocaloids when not used with Vocalochanger.

Voice direction on an AI can definitely result in similar-sounding voices, but those voicebanks tend to sound like real people who happen to have similar timbres; whereas the V6s have synthetic similarities I remember from previous versions of the engine.

 

[___]

In VOCALOID6 + DTM Station live stream it was said Megumi was directed to sing in a voice acted voice, not her natural one, meaning that V6 voicebank definitely contains data of the voice provider actually singing proper, there are also discrepancies, between the concatenative voicebank and the AI one to make it clear to me the two are working with different data ( will be specifically comparing Gumi V3 English to her V6 ), notably, Gumi V6 has slightly stronger japanese accent in English than Gumi V3 English and they also don't behave the same across the range and don't have same vocal ranges ( Gumi V3 English has has more solid low range whereas Gumi V6's isn't very good in lows but has stronger highs ).

The V6 is very faithful to her predeceasing voicebanks and she's definitely Gumi, but having worked more intimately with Gumi overall, V6 is definitely brand new recordings.

I'm also not sure about the sustainability of such method, where there would need to exist private standard-style vb for each V6 AI.

Then there's our own personal subjective anecdotal evidence, where we can just agree to disagree, because I personally do hear notable improvement in default naturalness.

 

[lIlI]

That's very interesting, it seems like not all AI is created equally - even on a single engine, some voicebanks sound more artificial than others. e.g. on SynthV there are voices that sound completely human, whereas some companies like AHS and AUDIOLOGIE have a Vocaloidy-ness to them. But without actually being on the ground floor making a voicebank, it's hard to know why.

 

[Granata]

Okay I've had this on my mind for a long time, time to let it out:

I don't want Rosa to come out at all. There are Shirayuki (her vp) sh*tacon accusations and from the screenshots I've seen it seems legit. Her Cevio barely got attention and I think the fandom would've been way more invested for an actual Cul update

Edit: Ik she isn't in Internet Co's priorities but I still don't want her

 

[AddictiveCUL (Add)]

Okay I've had this on my mind for a long time, time to let it out:

I don't want Rosa to come out at all. There are Shirayuki (her vp) sh*tacon accusations and from the screenshots I've seen it seems legit. Her Cevio barely got attention and I think the fandom would've been way more invested for an actual Cul update

Edit: Ik she isn't in Internet Co's priorities but I still don't want her

I agree a CUL's update would be way better than Rosa, but I wanted Rosa too T3T In an ideal world I would want they changeg the VP and release both (A CUL update and Rosa!)

 

[MillyAqualine]

Oh maaaaaaaan if it's true then it's pretty sad.... (Plus i was slightly rooting for ROSA... Seems I'll keep it to the other Cevio Rose)

 

[peaches2217]

Okay I've had this on my mind for a long time, time to let it out:

I don't want Rosa to come out at all. There are Shirayuki (her vp) sh*tacon accusations and from the screenshots I've seen it seems legit. Her Cevio barely got attention and I think the fandom would've been way more invested for an actual Cul update

Edit: Ik she isn't in Internet Co's priorities but I still don't want her

Do you mind sharing those sources? I'm not finding anything, and that's a hella serious accusation to make, so that's something I wanna know 100% for sure. (It's also worth noting that Internet doesn't own Rosa; they're just her distributer, so they don't directly dictate how and when she'll be updated, iirc.)

 

[Granata]

That was sent to a Discord server I'm in a good while ago. So she basically is sexualizing Len in a questionable way



In the art she's commenting he does look like a minor. And there's this song:

Song where she basically simps for him and the lyrics acknowledge he's 14

Idk it isn't that much but it's still weird to have someone like that being the vp of a commercial synth

Edit: Appearantly she also worked with Risumii on a song. For those who don't know Risumii is a producer who got arrested for being p*dophilic. That one's not that much of evidence but still makes her look weird

女子中学生とTwitterで連絡とり性行為か 容疑のボカロP「リスミー」逮捕 - ライブドアニュース

「リスミー」名義で作曲していた「ボカロプロデューサー（ボカロP）」。5月31日、京都府の青少年健全育成条例違反（淫行など）の疑いで逮捕された。女子中学生とTwitterで連絡をとり、京都市のホテルで性行為をしたとされる

news.livedoor.com

 

[Leon]

Len is not inherently or canonically a minor and a really big chunk of the fandom would have to be cancelled if thirsting over him was a problem.

That said, acknowledging that he's 14 in a song (so that, in that song he is in fact a minor) is creepy as hell. People are gross.

 

[Blue Of Mind]

Len is not inherently or canonically a minor and a really big chunk of the fandom would have to be cancelled if thirsting over him was a problem.

That said, acknowledging that he's 14 in a song (so that, in that song he is in fact a minor) is creepy as hell. People are gross.

I agree - under this logic, all the teenage fangirls in this fandom would have to be thrown under a bus for simping for Len, and let's not forget the people who thirst over Rin, Miku and countless other vocal synths who are either canonically or interpreted to be minors. Situation is still creepy though, I won't deny that, especially since the song acknowledges he's underage.

 

[Granata]

I'm really the last person who wants to force their Kagamine headcanons on others but well the song says Len's age, height, weight etc and at one point there's a shotacon mention so that weirded me out. And if you read the Tweet it's probably not satire

 

[WyndReed]

Edit: Appearantly she also worked with Risumii on a song. For those who don't know Risumii is a producer who got arrested for being p*dophilic. That one's not that much of evidence but still makes her look weird

Hey you got any links to proof about this? I’m subscribed to the musician, but can’t seem to find any info about this.

 

[Granata]

The article I sent says so and a few others talked about it if you want some Tweets

 

[WyndReed]

Ugh, that’s just gross. Thank you for posting though.

 

[kylin]

Okay I've had this on my mind for a long time, time to let it out:

I don't want Rosa to come out at all. There are Shirayuki (her vp) sh*tacon accusations and from the screenshots I've seen it seems legit. Her Cevio barely got attention and I think the fandom would've been way more invested for an actual Cul update

Edit: Ik she isn't in Internet Co's priorities but I still don't want her

Welp. Guess I'm never buying her lol. Thanks for sharing, but mannn that's dissapointing. I was interested in her voice, despite her sorta garbage design.

 

[AmazingStrange39]

That's very interesting, it seems like not all AI is created equally - even on a single engine, some voicebanks sound more artificial than others. e.g. on SynthV there are voices that sound completely human, whereas some companies like AHS and AUDIOLOGIE have a Vocaloidy-ness to them. But without actually being on the ground floor making a voicebank, it's hard to know why.

I've actually seen some people be impressed by some of the old concatenative Vocaloids as well. A big crowd-pleaser is AVANNA for example, and I personally also find Fukase to sound very realistic.

 

[junky]

Is it possible that Yamaha might still be using concatenative synthesis, with Vocalochanger being the only AI element? Or perhaps they're training the AI off concatenative voicebanks, to keep that classic Vocaloid sound - similar to people who create 'AI' versions of concatenative voicebanks by running them through Diff-SVC. I mention this because, by default, the V6s still sound very much like concatenative voicebanks. Fuiro sounds exactly how I would expect Philo's voice to translate to concatenative synthesis, at least from what we've heard so far, and the English V6s have the same pronunciation quirks as previous English Vocaloids when not used with Vocalochanger.

Voice direction on an AI can definitely result in similar-sounding voices, but those voicebanks tend to sound like real people who happen to have similar timbres; whereas the V6s have synthetic similarities I remember from previous versions of the engine.

I agree with this, said this prior and people disagreed lol
V6s sound A LOT like pre-V6 voices to me..meanwhile NNSVS (cuz that's basically the UTAU equiv to AI banks right?) banks sound way different than their VPs UTAU
Plus, Vocaloids who've moved to SynthV also increase in quality tenfold, I bet that's why a bunch of people want Miku to go to SynthV (even if it's super unrealistic and has a 90% chance of never happening, pretty sure CFM solely gave permission and directed Teto SV and didn't actually help create the bank) but V6 AI still sounds..Vocaloid-y, lol
See: The difference between Gumi V2/V4 & V6 vs. the difference between Stardust & Haruno Sora's Vocaloid and SynthV .. SynthV Stardust doesn't even sound like her to me (guess that's why they went with a different name, Infinity)

2 of the ZOLAs not even needing re-recording shows it at least *is* possible to make an AI bank from a concatenative bank, even with Talk banks
(on an unrelated note: I'm so surprised happy ZOLA got updatespretty sure no one saw that coming especially their talk banks. Along with Porterloid. V6 might be my favorite engine so far, wish I had money)

 

[SilSinn9821]

My own unpopular opinion: VOICEMITH had no need to age up Xia Yu Yao’s canon age from 17 to 18 in June 2016 (months after they had released her Mandarin 3.0 UTAU voicebanks). Her lore never had any Fee-chan-like controversies of its own, plus her usage guidelines always prohibited (& still prohibit, even after the age-up) any R-18/NSFW works; also, her merch never included the kind of questionable cringe that SSS LLC (Zunko et al.) is known for on its own BOOTH shop. Had her age been kept at 17 for her Synthesizer V AI voice database, her reception would’ve been no different than the reception Haruno Sora & most other canonically-17-year-old VOCALOIDs & SynthVs ever got (& still get).

 

[Granata]

I mean was it neccessary? No, but maybe they aged her up as a symbol for how her voicebank evolved. So that the UTAU is a child who has grown into an adult (SynthV). Ofc that's just my theory

 

[WyndReed]

I feel like the Synth V AI engine has changed enough since Anri was originally released that Anri Arcane being a separate voicebank, rather than just an update, might be a good thing. I’m not super fond of updating the software/voicebanks and then having them sound different or have different glitches/bugs. (I know that yamaha never updating vocaloid and then just changing people more money for a newer engine is a big sticking point in the community, but the synth v engine updating definitely throws me for a loop sometimes)