• We're currently having issues with our e-mail system. Anything requiring e-mail validation (2FA, forgotten passwords, etc.) requires to be changed manually at the moment. Please reach out via the Contact Us form if you require any assistance.

Other Maghni AI by VocaTone Studio and Misbah Studios

Prism

Enthusiast
Jul 18, 2019
524
I think the accent feature is a great idea they mainly talked about it for non native singers. I do hope it is a feature that the user has control over, because I could see some people really liking accented vocals like say macne nana while other finding her hard to use or hard to understand.
 
  • Like
Reactions: Alphonse

___

Oct 8, 2019
1,546
I think the accent feature is a great idea they mainly talked about it for non native singers. I do hope it is a feature that the user has control over, because I could see some people really liking accented vocals like say macne nana while other finding her hard to use or hard to understand.
For me personally that's on the iffy side. I think what we should be doing is encouraging ( now let's say for japanese accent but applies for any non-native accent ) japanese people, well versed in english, to be recording eng jp accented voicebanks instead of forcing those not comfortable with it to record them and letting native and fluent english speakers on their develpoment teams when developing english voicebanks to ensure good quality but I am not sure whether erasure is the right approach.

It's just that in this community there's this mindset that non native accents are inferior and should be shamefully masked and covered up and I think it's one that shouldn't be perpetuated.

As for the understandability and how hard it is to use, there are native english dialects and accents that another native english speaker wouldn't understand and would most likely have hard time using in syntehtized form simply because they're from another part of the world, american might have hard time using scottish accented synth, but....it wouldn't be fair to suggest that those accents should be covered up as much as possible and the same should apply to non native accents.

No hard feelings just explaining where the distaste for such feature might be coming from!
 

Prism

Enthusiast
Jul 18, 2019
524
I would like to see what maghni Ai considers as an accent like tone or pronunciation they seem to be wanting to offer a lot of different languages so I can see this helping their goal. If the user could edit the accents from a list of accents It would add to their flexibility. We know the spanish community really wants a non european accented comercial vocalsynth so it would be nice if they offered different accents to morph it into so everyone can finally have a synth in their accent. I'm not saying accents are "shameful". What is normal speech to one person is heavily accented to another. I am saying that we should have the flexibility and option to edit the accent so anyone can understand them and not having to manually edit them bits to have them sound understand to certain ears making them more use able to certain markets.

also no hard feelings on my side too
 

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma
I would like to see what maghni Ai considers as an accent like tone or pronunciation they seem to be wanting to offer a lot of different languages so I can see this helping their goal. If the user could edit the accents from a list of accents It would add to their flexibility. We know the spanish community really wants a non european accented comercial vocalsynth so it would be nice if they offered different accents to morph it into so everyone can finally have a synth in their accent. I'm not saying accents are "shameful". What is normal speech to one person is heavily accented to another. I am saying that we should have the flexibility and option to edit the accent so anyone can understand them and not having to manually edit them bits to have them sound understand to certain ears making them more use able to certain markets.

also no hard feelings on my side too
I think the answer here isn't the ability to distort accents, it's making more VBs with varied accents and singing styles. That's a bridge that will need to be crossed if they get there, but like... to me, the thought of distorting a VB's accent or style is kinda like trying to make Sachiko sing Idol Pop because you don't like an Enka-style voice. It just doesn't feel right.

EDIT: Ah, to clarify! It's like manipulating Sachiko to sing Idol Pop instead of using a VB that's suited to it, not for the challenge, but simply because you don't like the way she sounds by default. Like... like a more extreme version of the whole Sakine Meiko situation.
 
Last edited:

___

Oct 8, 2019
1,546
I would like to see what maghni Ai considers as an accent like tone or pronunciation they seem to be wanting to offer a lot of different languages so I can see this helping their goal. If the user could edit the accents from a list of accents It would add to their flexibility. We know the spanish community really wants a non european accented comercial vocalsynth so it would be nice if they offered different accents to morph it into so everyone can finally have a synth in their accent. I'm not saying accents are "shameful". What is normal speech to one person is heavily accented to another. I am saying that we should have the flexibility and option to edit the accent so anyone can understand them and not having to manually edit them bits to have them sound understand to certain ears making them more use able to certain markets.

also no hard feelings on my side too
Yes I definitely see your point and the positives you brought up are 100% true and valid, I just feel like there two sides to one coin and while you yourself aren't malicious there are users like that and it could backfire and it's worth wondering whether it's worth it when we could be giving opportunities to people of those accents/dialects instead.
 
  • Like
Reactions: peaches2217

Prism

Enthusiast
Jul 18, 2019
524
Of course, we got to wait till they actually make the software to hear how it sounds. I'm just saying it as flexibility thing we don't get that many non japanese non chinese comercial vocal synths so having the ability adds a lot to the usability. Also of course if they get demand for a certain accent they should get a voicer for that accent. We haven't gotten a english voice since 2018 with elanor if you don't count emvoice and we haven't gotten a Spanish bank since 2013. So it is important to have these banks be versatile.
 
Last edited:

___

Oct 8, 2019
1,546
Ultimately, it was more complex discussion, on one hand there is flexiblity, on the other is erasure and what to take from this is not to necessarily decide which one overweights the other but to show the developers to not so recklessly jump onto the first idea they have, they're little too easily convinced and they should start more critically thinking about what they implement.
 

Leon

AKA missy20201 (Elliot)
Apr 8, 2018
979
As far as a non-native speaker voicing a bank goes, the way I see it is that an accent can be cute and fine, but only to an extent. For example, I find Nana's English accent pretty cute because overall (with a good deal of phoneme editing) she can be understandable. However, Miku English is a lot harder to make understandable, so people had a lot of complaints about her V3, and the V4 is better. Native Chinese speakers also complained about Miku Chinese until Crypton released a patch to help it. If it's a product -- an instrument -- that someone is paying $100 (give or take) for, I think not being able to understand it is a valid complaint.

In any case I don't really think accent erasure is going to happen so much as, they're attempting to alleviate the issue of Latin American vs. Spanish accents because it can do both so the user can choose, or how in V1 the vowels tended to be very "British" and if I'm using Leon and I say "oh gross that vowel sounds off if I tune the rest of his phonemes to sound how I'd sing it ie. vaguely American", I can go into the Resonance parameters and change the vowel sound how I like. It's a matter of convenience and pleasing the largest audience possible, and if they can pull it off then I think it's a smart move. Certainly, there can be unforeseen consequences, but I don't think it will end up being a huge deal 🤔

But, we shall see!
 

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma
As far as a non-native speaker voicing a bank goes, the way I see it is that an accent can be cute and fine, but only to an extent. For example, I find Nana's English accent pretty cute because overall (with a good deal of phoneme editing) she can be understandable. However, Miku English is a lot harder to make understandable, so people had a lot of complaints about her V3, and the V4 is better. Native Chinese speakers also complained about Miku Chinese until Crypton released a patch to help it. If it's a product -- an instrument -- that someone is paying $100 (give or take) for, I think not being able to understand it is a valid complaint.

In any case I don't really think accent erasure is going to happen so much as, they're attempting to alleviate the issue of Latin American vs. Spanish accents because it can do both so the user can choose, or how in V1 the vowels tended to be very "British" and if I'm using Leon and I say "oh gross that vowel sounds off if I tune the rest of his phonemes to sound how I'd sing it ie. vaguely American", I can go into the Resonance parameters and change the vowel sound how I like. It's a matter of convenience and pleasing the largest audience possible, and if they can pull it off then I think it's a smart move. Certainly, there can be unforeseen consequences, but I don't think it will end up being a huge deal 🤔

But, we shall see!
It should be taken into account that Macne Nana's voice provider is fluent in English while Miku's is not; Miku's V4 English is better than her V3 English because they didn't try masking her accent as intensely, which resulted in better more natural sounding product. And from what I've heard, the issue with her CHN wasn't necessarily the accent as much as general muffled-ness and fuzzy engine noise. She had accent issues, yes, because again Saki Fujita doesn’t speak the language she was recording, but that was hardly the predominating issue. The point would be to hire voice providers comfortable with the lamguage instead of forcing those unfamiliar with it, I think.

I don’t personally take issue with an accent being pulled back when it’s genuinely non-understandable to most speakers. The issue is when the accent is perfectly understandable and someone still wants to change it simply because it doesn’t sound “right”. It’s not outright offensive, but it definitely doesn’t sit right is all. Again, that just goes back to having VPs who are fluent in, or at least intimately familiar with, the language they’re recording.

Forgive the soap box!! Just wanted to explain that particular side of the discussion.
 

Leon

AKA missy20201 (Elliot)
Apr 8, 2018
979
Totally agreed about the VP, but I think that's where some companies are smart in attempting to match voices while changing voice providers for different languages! There's obv pros and cons of that method, but if you're a company hoping to get a product to sell then it needs to be decent quality, so the tradeoff may be worth it to you... As a listener I'd rather the tone maybe be slightly different because of a different VA than for it to technically be understandable but sound really "stiff" because the VA wasn't comfortable in that language, type thing. But that's personal preference that I understand plenty would disagree with! There's no clear answer here, of course. Situational and all. It's just something to think about :)

I do think you're right about Miku CHN's muffled-ness being the biggest issue, actually-- although I definitely have heard a good amount of trash talk about her accent 😭😅
 

Blue Of Mind

The world that I do not know...
Apr 8, 2018
699
For a while, I didn't realise Tianyi had a different VP fill in some parts of her Japanese VB because the tones between her Japanese and Chinese banks are fairly similar. So long as they don't sound drastically different, using a different VP to help with a separate language bank isn't a problem.

(Going off-topic, the main reason why I hate IA's Talk bank is because they used a different VP other than Lia, which meant that the two banks ended up sounding completely different from each other and it's pretty obvious that it's the case).
 
Last edited:

___

Oct 8, 2019
1,546
I and my girlfriend talked about this yesterday and now I am little confused...could someone clear this up? Chances are they might have talked about this but I've forgotten since there's been a lot that's been said or it could be that my understanding of AI banks is just superficial but...

What exactly justifies them having the AI title when it's sounding like they're making sample based voicebanks? From talking about reclist when....from my understanding, that isn't really a thing for AI banks? Or how they talked about "training voicebanks" which is just getting the vps ready for the legit recording process, getting used to the whole environment, reclist etc. when....AI banks are made by examining singing data of the voice provider? What exactly would be there to preapapre for? Like....Neutrino Kiritan....didn't they have just have recordings of her vp singing on hand and wen't "sure why not lol"?
 

Prism

Enthusiast
Jul 18, 2019
524
Okay so what they are planning on is having an ai auto tune button with a sample based library that's it from what I heard. I do find ai in the name to be misleading. Yeah kiritan wasn't made for neutrino she was just a singing dataset(wav files of plain singing and the score) for research that anyone can look at and use. I do think that is one huge bonus for ai is that the talent can just sing and not have to worry about phonemes and can emote
 

Silver1063

Aspiring Fan
Apr 19, 2019
52
I think other/(all?) parameters will use AI suggestions as well, whatever that will look like. Omg what if they use the same AI model for every voice....
 
  • Like
Reactions: ___

Prism

Enthusiast
Jul 18, 2019
524
It would be okay in my opinion but not desirable. I would be okay if it was like 5 with different styles like belting or jazz singing
 

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma



If I may... I've noticed all of their updates since beginning work at the beginning of the month have been focused on building reclists and introducing "capabilities" for making languages... it all still feels kinda like they're doing everything they can to avoid, y'know, getting the synth and their two announced VBs built. Is it because making reclists is free and they'd still rather not put their own funds into building a basic version of their final product? Are they just so distracted by all these cool ideas that they're focusing on that, rather than on getting their actual product out there for potential funders to see? I don't know, but it doesn't sit right with me.

I get that they weren't planning on using their own funds to make it initially, but if money is an issue, I feel like they should be focusing on making that happen by other means rather than spending all this time making lists that have "capabilities" for all these languages. Because at the end of the day, what good is any of that if they still don't have an engine to put it on? Six months (coming up on five) isn't a lot of time to build an editor and two voicebanks, even if they're just stripped down for the sake of demonstration. Unless they're doing some heavy work behind the scenes... it sounds mean, I know, I'm sorry, but it feels like they're twiddling their thumbs and wasting time.

Peaches now steps down from her soapbox. Forgive me. I've been hoping for the best, but it reads like they're not doing anything differently than they were in December.
 

Rolo

Watch Levius and Stan Natalia Cromwell
Sep 16, 2019
195
At this point I really want to hear what the VA’s sound like even if that’s just a raw singing demo (no software) because it’s just weird we still don’t know what the voices will even be like. It would take nothing to quickly record that.
 

Users Who Are Viewing This Thread (Users: 0, Guests: 1)