• 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.

SynthV CeVIO AI / Synthesizer V Tsurumaki Maki

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma
The segment at about 1:18:00 started by him saying "this is CeVIO AI Maki", then he proceeds to play her JP voices followed by her ENG voices. It seems the entire episode was centered around CeVIO/Talk banks only.
Sorry. I’m at work so I can’t watch the whole thing right now. The Cranky Manager is on duty so just watching what little I did got me a stern lecture about the importance of a single-minded focus on unloading freight.

Already people are complaining about her accent as I expected 😅
I’m disappointed but not surprised. If the English-speaking fandom is good at one thing, it’s complaining. :clara_ani_lili: I really do hope more people warm up to her! She sounds really nice so far.
 

Rylitah

kiyoteru enthusiast
Staff member
Moderator
Apr 8, 2018
575
She sounds wonderful!!!! I couldn't have asked for a better first English Talk Synth.

My only disappointment lies with her JP Talk actually, because.... yeah, I can really tell she has a new voice provider since that sounds nothing like Maki's original voice. I'm not that bothered though, I own her Voiceroid anyway, so I can use that one whenever!

I did see some complaints but I think it's mainly just about how awkwardly chopped up the sentences are? Which I'm sure you can iron out with some editing to make her sound more natural.

Waaaagh I'm so excitedddd
 

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
I'd say a majority of people commenting/retweeting my tweet with the clip are like "SHE IS SOOO GOOD?!" and I'm not exaggerating. The rare complaint is that she is not the quality you would expect for an expensive AI bank.

I'm kind of annoyed by some people saying people are complaining about the accent. It's not just that! I think if you look at it from an outsider's point of view, it's like this:

Pros:
  • Maki sounds cute!
  • Finally an English TTS!
  • Can't wait to hear her Synth V English!
  • It sounds like her VP because it's AI > Faithful reproduction
  • TTS is helpful for having speaking parts in Maki songs or little skit videos
  • Her voice has emotions!
  • She has novelty similar to Adachi Rei aka being a pioneer

Cons:
  • It sounds like her VP because it's AI > The English in that clip has parts that are very unintelligible. If you say it's 100% clear/understandable for every word she said, I do not believe you are being honest. Go ahead and write down what she said for me without having to replay it several times.
  • Yes, you can edit her pitch/intonation/speed in the editor. But this is an AI BANK. You shouldn't HAVE to edit the phonemes that much, it's AI for a reason (saving time with tuning natural speaking + emotion capabilities). I argue that with an AI bank, there should be minimal editing. I have messed around with CeVIO, VOICEROID, and A.I.VOICE and it is very time-consuming to edit pronunciation by hand. I am not impressed to say the least.
  • TTS is helpful for having speaking parts in Maki songs or little skit videos > Are you going to look me in the eyes and say that Maki English is on par with FREE English TTS (aside from the perk of her having emotions) in terms of understandability without subtitles? I can definitely see her being used in skits or videos aimed at fellow vocal synth lovers or used by Japanese users who like English. Can I see it being used for narration of homework projects or YouTube videos aimed at normies/people who don't know about synths/people that are willing to listen to accented English? No. I can see it being mocked.
  • I can almost guarantee anything A.I.VOICE releases in December for English is going to be of more usefulness than Maki's English TTS voice, because they actually have tons of examples of their multilingual voices. Looking back on Maki is probably not going to be impressive in terms of longevity.

Note: I do NOT hate accents on real people. And I love hearing English learners speak English no matter how skillful. However, this TTS bank is going to cost about $160. Yes, it's useful for Maki fans who want to make skits/speaking parts in songs. But no, I absolutely do not agree that this is a quality voice bank for native English speakers to make narration aimed at other native speakers. My viewpoint is purely from a product standpoint and not a character bank standpoint. Yeah, it's "fine" for what you'd expect from Maki but it is in no way "SOOO GOOD", there are definitely legitimate problems with this product.
 
Last edited:

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma
God I can’t wait to hear her song bank in its full glory... I never really knew it got attached to Maki as a Voiceroid, so the different VP doesn’t bother me. I still wish they’d gotten her OG VP for the sake of Voiceroid fans, but as she is, from someone who’s just now getting familiar with her? She’s amazing :ia_ani_lili:
 

___

Oct 8, 2019
1,546
Also I just wanna say! We all have different experiences, what is understandable for one person might not be for another, one person might have harder time with understanding one accent than the other, I know I do and that's fine, but I understand Maki just fine and it's not universal experience that she's unintelligible and it's not right to say people are lying because your experience differs :one_smile_lili:

However I understand that makes her more niche and not as flexible and it's fair to say or wish that the 1st TTS of such kind would've been more...universal :teto_lili:

No hard feelings I just wanted to say that :akasakiminato_lili:
 
  • Like
Reactions: peaches2217

VocAddict

The Voice Within Us
May 8, 2018
55
24
The Other Side
voxil.net
Cons:
  • It sounds like her VP because it's AI > The English in that clip has parts that are very unintelligible. If you say it's 100% clear/understandable for every word she said, I do not believe you are being honest. Go ahead and write down what she said for me without having to replay it several times.
  • Yes, you can edit her pitch/intonation/speed in the editor. But this is an AI BANK. You shouldn't HAVE to edit the phonemes that much, it's AI for a reason (saving time with tuning natural speaking + emotion capabilities). I argue that with an AI bank, there should be minimal editing. I have messed around with CeVIO, VOICEROID, and A.I.VOICE and it is very time-consuming to edit pronunciation by hand. I am not impressed to say the least.
  • TTS is helpful for having speaking parts in Maki songs or little skit videos > Are you going to look me in the eyes and say that Maki English is on par with FREE English TTS (aside from the perk of her having emotions) in terms of understandability without subtitles? I can definitely see her being used in skits or videos aimed at fellow vocal synth lovers or used by Japanese users who like English. Can I see it being used for narration of homework projects or YouTube videos aimed at normies/people who don't know about synths/people that are willing to listen to accented English? No. I can see it being mocked.
  • I can almost guarantee anything A.I.VOICE releases in December for English is going to be of more usefulness than Maki's English TTS voice, because they actually have tons of examples of their multilingual voices. Looking back on Maki is probably not going to be impressive in terms of longevity.

Note: I do NOT hate accents on real people. And I love hearing English learners speak English no matter how skillful. However, this TTS bank is going to cost about $160. Yes, it's useful for Maki fans who want to make skits/speaking parts in songs. But no, I absolutely do not agree that this is a quality voice bank for native English speakers to make narration aimed at other native speakers. My viewpoint is purely from a product standpoint and not a character bank standpoint. Yeah, it's "fine" for what you'd expect from Maki but it is in no way "SOOO GOOD", there are definitely legitimate problems with this product.
I understood most of what she said on my first listen which is pretty awesome to be honest since 1) she's a native Japanese speaker, 2) it's an AI bank and 3) the demo was done by Japanese people. I really do not know what you expect from something that is making use of neural training to produce a bank?? It will not sound different from its source, and that's the intention.

Just because it is an "AI" bank doesn't mean that you don't have to do any work. We've already seen how Saki's AI bank works when many people covered the same song and you end up getting the same output. All AI anything does is give you a base to work from, if you want to change how it sounds, you have to put in the work. You got to remember that the 'A' in 'AI' stands for artificial and not autonomous.

It's Maki for goodness sakes. AHS is not catering this to the average English joe for voiceover on his Minecraft videos. This is Maki, a Japanese voice that just so happens to also have an English one. They're buying because she's Maki. They catering to a specific niche, and there's nothing wrong with that. Eventually we might get another voice from them that is more suited for general purpose use, maybe not but the point still stands that this is the product that was made.

If you want something for native English speakers, get a native English company to make a bank for you or a company that has multilingual experience. Don't expect from a company that has been making Japanese banks for their entire existence to be "oh so amazing" with their first product. We've all seen how most English Vocaloids sound from Japanese VPs, there's no reason to expect it to be different for TTS, especially when it's AI trained. It's clearly not aimed for you.
Funnily enough, while she doesn't sound American, I don't hear much of a Japanese accent either. I saw someone say she sounded British, so perhaps that's it?
Yeah, it seems that she learned non-rhotic English, so essentially British in essence. I think it's a nice quirk when everyone is trying to sound American these days.
 

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma
Exactly. Not everyone speaks english with an northeastern american/toronto area canadian accent.
Fun fact: being from the South, unless something is in perfect Southern American English, I can’t understand it. Reckon that ole’ buncha doohickey goes right over m’head, y’all hear?

Okay okay, joking aside: she’s so unlike other English banks we’ve had so far, at least in terms of speaking banks. Her accent is obviously but subtly Japanese, and I can’t quite put a word to the rest of its sound. But it’s so clear and lovely and I can see myself using it a lot, at least her English song bank.
 
  • Haha
Reactions: ___

Prism

Enthusiast
Jul 18, 2019
524
Honestly she's on par or a little less than the my little pony tts maybe a little less noise. If you can edit the phonemes that she'll be okay but her accent is strong and sounds like her voice provider. It is going to be a hard sell for the mainstream english market though. I have seen the mlp tts used in skits so she might have a chance but the accent and price will be a hard sell
 

peaches2217

Give me Gackpoid AI or give me DEATH
Sep 11, 2019
1,930
26
Arklahoma
Honestly she's on par or a little less than the my little pony tts maybe a little less noise. If you can edit the phonemes that she'll be okay but her accent is strong and sounds like her voice provider. It is going to be a hard sell for the mainstream english market though. I have seen the mlp tts used in skits so she might have a chance but the accent and price will be a hard sell
As VocAddict pointed out, she’s not marketed to a mainstream English audience, so I don’t think the accent’s going to be an issue to her target audience. And honestly, I don’t even think her accent is that thick; it’s very clearly not native, but it’s perfectly understandable.

This DOES make me hopeful for potential future products aimed at a more mainstream English market though! Imagine this quality with a native English speaker. It would top anything else we have right now.
 

___

Oct 8, 2019
1,546
Yeah I expected people being divisive about the accent, she's not for everyone in the western community and that's fine. Personally I think it has it's charm...ngl it sounds a lot like the voices some Vtubers put on :teto_lili:
 

uncreepy

👵Escaped from the retirement home
Apr 9, 2018
1,618
Just because it is an "AI" bank doesn't mean that you don't have to do any work. We've already seen how Saki's AI bank works when many people covered the same song and you end up getting the same output. All AI anything does is give you a base to work from, if you want to change how it sounds, you have to put in the work. You got to remember that the 'A' in 'AI' stands for artificial and not autonomous.

It's Maki for goodness sakes. AHS is not catering this to the average English joe for voiceover on his Minecraft videos. This is Maki, a Japanese voice that just so happens to also have an English one. They're buying because she's Maki. They catering to a specific niche, and there's nothing wrong with that. Eventually we might get another voice from them that is more suited for general purpose use, maybe not but the point still stands that this is the product that was made.

If you want something for native English speakers, get a native English company to make a bank for you or a company that has multilingual experience. Don't expect from a company that has been making Japanese banks for their entire existence to be "oh so amazing" with their first product. We've all seen how most English Vocaloids sound from Japanese VPs, there's no reason to expect it to be different for TTS, especially when it's AI trained. It's clearly not aimed for you.

Yeah, it seems that she learned non-rhotic English, so essentially British in essence. I think it's a nice quirk when everyone is trying to sound American these days.
I wrote an apology on Twitter for writing my humble opinion in an angry tone. I shouldn't have written it so passionately/assumed people would agree. But I have legitimate reasons Maki is not up to par for me personally.

I use TTS many hours a week, I can not read long passages without it in English or Japanese. I own CeVIO, VOICEROID, and A.I.VOICE for Japanese and use Microsoft TTS for English. When Maki got announced, I assumed she would be of quality similar to what AI Inc has to offer (on this page, scroll down for audio clips 様々な言語での音声合成を可能にします。 AITalk International® )

It was dumb to assume that I would get a certain quality, but at the same time, I REALLY wanted a TTS of a character voice that could narrate to me instead of Microsoft Mark.

When I use the CeVIO/VOICEROID/A.I.VOICE banks, I input the text and do not have to edit it much. It's just type and export, I do not want to tune speech that is supposed to sound natural by default, especially through AI. I don't have to tweak the voices most of the time in VOICEROID/A.I.VOICE and I perceive Maki as needing excessive tweaking that isn't worth the effort based on the sheer amount of TTS narration I need per week. I do not appreciate being told by people (ex: Twitter) that I am using TTS wrong by assuming I could've used it for educational narration when they think it's clearly meant only for skits or that I'm "using it wrong" when I do spend many, many hours listening to TTS and I know what I want/need from a voice bank.

Also, I legitimately thought that Maki was not 100% understandable but people got really angry at me because they said they can understand her fully. For me, there is no purpose in owning a voice bank I can not understand just by listening (because I can't pay attention/repeatedly lose my place while trying to follow along to text, I NEED to be able to understand the voice without seeing the words). She might be the perfect character voice for some people, but I was planning on having a voice to help me do my daily function of reading + use for my educational website. I got worked up because Maki isn't how I imagined she would be based on my experience with other TTS.

Lastly, I am very uncomfortable with people accusing me of disliking Japanese accents. I have been speaking in Japanese and English to bilingual Japanese users on Skype since 2008. I am used to accents and like it when people speak English with whatever ability they have. But this is a product and not a person per-say in my humble opinion. My criticisms are purely because of my perceived time to edit her voice and having some words I can not understand being said, it's about my accessibility.

We won't know her true quality until a demo page comes out and we can purchase her. And I think we need to realize that TTS has fans that are both people collecting unique/novelty voices as well as people needing TTS for actual accessibility. It's like how some English song banks are hard to use due to wonked up phonemes even if it's a native voice bank or not.
 

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