Unpopular Opinions

[mobius017]

I agree with a number of the points made here.

On the one hand, in working with Miku English V4, I've personally found her pronunciations quite understandable (though not always perfectly so) when I've just put the words in and let the dictionary do its work. She's good enough, in my mind, that in listening to some of the songs where she struggles, I wonder how it's happening. I think it may be due to a number of potential causes:

1. Possibly the producer of said song was using Miku V3 or Engrish. IMO, V4 does a much better job than V3, although V3 can be invaluable for providing a darker tone via XSY.
2. Since I'm putting the words in during my usage, I know what the words are already, so it's harder for me to judge fairly.
3. It's the work with the parameters that make the words in the songs less understandable. I think it's possible that there is a "best practice" workflow that produces better results. In my theorization, this is based on a general rule of making the biggest, most sweeping changes through effects, plugins, and DAW automation, and smaller changes via the vocal synth parameters. I think if you work with the parameters first and push them too hard, you may end up making a sound that's more distorted than artistically improved.
4. Possibly the mixing/conflicts between the vocals and the background are to blame.
5. XSY can cause weird issues that would seem difficult/impossible to fix.

At the same time, I would never discount the merits of phoneme replacement. If it is your wish to change some of Miku's less native pronunciations, phoneme replacement is going to be a go-to technique. Lupin/nostraightanswer does fantastic English with Miku (Check out "On Top" and "Someday You'll Notice The Evening Colored Horizon (feat. Hatsune Miku)," and he credits lots of phoneme replacement for his results.

I don't think either approach is wrong. I'm personally pretty happy with what Miku does on her own, though as I have the need I will probably hope to learn to use phoneme replacement to help her pronunciations more than I currently do. Having said that, I think criticism against her bank could reach the point of being unfair if the complaint is that she doesn't have ultra smooth, native pronunciation. Just like Saki Fujita, she's a Japanese woman singing in English. That's a steeper hill to climb for her and for you as a producer than if you were working with a native Engloid. It's definitely doable--look at Lupin's songs, or "Thousand Little Voices" by Vault Kid and Flanger Moose--but it's more work.

 

[peaches2217]

I don't think phoneme edits are a bad thing by any means, and are really useful in cases like swapping [t] for a [d] or something for smoothness' sake and for the occasiona vowel swap! I think it has to be said though, that the more intricate and complex phoneme editing tends to only make a significant difference for the user tuning them, and rarely will they make a big enough difference to be impactful in the final piece. What I mean to say is... There should be a balance between relying heavily on phoneme editing and also letting the mix do some of the work for you, while also considering "would the average listener even notice that I changed this one tiny phoneme?" I do the same with my tuning and it's immensely helpful in reducing both my workload and improving the quality of my work.

Maybe this is the proper unpopular opinion i've been getting at, but I think Engloid stans have ended up conditioning themselves to believe Engloids are inherently bad at pronunciation, thus not bothering to pay attention to pronunciation, or even recognise the pronunciation as good to begin with. Just because a voicebank isn't recorded in an American or British accent doesn't necessarily make it horrible at pronunciation, and is quite disheartening to hear when Saki Fujita and Haruna Ikezawa likely worked their ass off when it comes to pronunciation. Having an accent that may be harder to understand to native English speakers isn't grounds to devalue a product and never should be.

In going on that, I think a lot of people are heavily oversimplifying the issue at heart and equating it to "Engloid pronounce word bad". English is a deeply complex language that's difficult to replicate in synthesis. You can edit every last phoneme to your heart's content and an Engloid will never sound human, just closer and closer to it. That said, the only Engloid I've ever worked with whose pronunciation was genuinely bad, not just unnatural, was Eleanor Forte. I gushed and oohed and aahed over her when I posted my first cover featuring her, but 80% of the work I put into that cover was fixing her pronunciations, because she would say the most basic words just outright wrong ("decide" she initially said as "deh-see-deh" - her pronunciation's a lot better in her SVS port thankfully). But she continues to be praised as the Gold Standard of English Synths.

So there's a lot more to English vocalsynths than their pronunciation that determines if they're seen as good or bad; however, the minute one has a flow that sounds less than human, they're slapped with the Bad Pronunciation label, because that's just the easiest and most convenient label to pull out.

 

[Mika]

I'm just gonna admit here right now, that as an non-native english speaker, nowadays I barely understand anything that is sung in english/american songs. I've heard a lot of people say they feel like Vocaloids over-pronounce words while singing, but to me real people currently are under pronouncing everything. It just sounds so muffled to me all the time that I just really do not care about vocal synth pronounciation anymore. As long as it sounds nice, who cares, I can't understand these songs just by listening anyways, no matter if a human or synth sings them lol. Idk why this became a trend because I have no issue understanding songs from 10-20 years ago but new stuff? Nope. I mean yeah, being able to perfectly understand a synth vocal is really cool and impressive but. If that isn't even the goal anymore for real people, why should it be for synths.

 

[___]

Privately, I've expressed how I don't mind not understanding every single word an english synth sings, it's something not every human singer does why should I hold english synths to a different standard? I personally value enunciation over pronunciation in english synths.

Perfect pronunciation is held as the golden standard for english synthesis but isn't it really just a preference at the end of the day? There are synths who are actively taken in such direction and it's okay to want that from engloids but why should be synths who don't choose to do so be given such a hard time, again, they should be held to a certain standard but as long as they're no Sonika I see no issue.

It goes back to what I was saying "there comes a point where it's not the product that's at fault but, imo, you for having unreasonable expectations from it". Cyber Diva doesn't appeal to me, I'd take Nana over her, but I don't give her hard time for taking more effort to be made expressive, that's why I haven't bought her and if I did, I would be aware of what I'd be getting myself into.

 

[Leon]

Under-pronouncing things is how English is often spoken and sang, but I've come across this in foreign languages too. So although it's frustrating to understand non-native languages (believe me, as a fan of learning other languages but as someone who also sometimes has listening comprehension problems, I totally get the pain) I think that the goal shouldn't necessarily be to make a synth say words as "correctly" as possible. If I say "natural" in this sense, I mean that I'd like something to sound like how someone would normally say it, not that I want the bank to sound 100% totally human and not obviously a computer. (Sure, humanlike synth is impressive, but sounding synthesized can be great too.)

I do think there is a difference between someone complaining that a sound doesn't come across correctly (re: the Miku and Nana complaints about how they often straight up can't say a letter) and actually mocking an accent. I'm on the side of, I'm allowed to critique this $100+ instrument I've just bought. That being said, the line can be easy to cross if someone isn't careful, and I don't think a heavy accent actually ruins a bank or anything. Nana English V3 is actually one of my favorite English banks. She's wonderful.

I actually am one of those overly neurotic people who phoneme edit probably every single word with Engloids. I want them to say things how I'd say them, approximately. I know several others who edit even more than I do, not all of them native English speakers (but fluent). I'll fully admit we make it harder on ourselves than we need to

And to end it all, I'll also say that for every person I've seen complain about non-native English banks' pronunciation, I've seen someone praise those banks for being cuter and dunking on the native Engloids. There're going to be debbie downers in every corner of a fandom, and I find it best to just kinda... observe from the sidelines and stick to my own corner most of the time, LOL

 

[Scarlet Illusion]

I do think there is a difference between someone complaining that a sound doesn't come across correctly (re: the Miku and Nana complaints about how they often straight up can't say a letter) and actually mocking an accent. I'm on the side of, I'm allowed to critique this $100+ instrument I've just bought. That being said, the line can be easy to cross if someone isn't careful, and I don't think a heavy accent actually ruins a bank or anything. Nana English V3 is actually one of my favorite English banks. She's wonderful.

Yeah, I agree wholeheartedly with this! Like, there's a lot of non-native English singers I do like (see: the English songs in the Attack on Titan soundtrack, for example), but sometimes when sounds don't get pronounced correctly, it can be jarring for me--particularly "r"s being pronounced with a "w" sound instead; it's jarring because it creates a bit of a disconnect since you're expecting to hear one thing, but hear another. I'd akin it to when you're trying to sing a song and you think you know the lyrics but get thrown for a loop because the last chorus changes up the lyrics. I know it's not the best example, but it's the best I've got.

I think it's fine to have your qualms with the pronunciation, but, of course, it's a totally different thing to blatantly make fun of it. "I like this singer but the pronunciation could be a little better at this part; they could do it more like {explanation} instead of {whatever is being critiqued}" vs "LOL they can't pronounce 'r', that so weird!!! They talk like a baby, eks dee!"--one is constructive while the other is just plain rude.

 

[Scarlet Illusion]

I'm not a huge fan of Pinocchio-P's music. Don't get me wrong, I'm not saying he's bad by any means, but his songs just feel very plain to me and lack something to really draw me in. In other words, most of them are just kind of boring.

 

[peaches2217]

I'm not a huge fan of Pinocchio-P's music. Don't get me wrong, I'm not saying he's bad by any means, but his songs just feel very plain to me and lack something to really draw me in. In other words, most of them are just kind of boring.

I remember listening to Slow Motion and thinking the whole time “Okay, so when does it kick into gear?” No shame to him or any of his fans! All his stuff just kinda blends together for me.

 

[morrysillusion]

I remember listening to Slow Motion and thinking the whole time “Okay, so when does it kick into gear?” No shame to him or any of his fans! All his stuff just kinda blends together for me.

its funny you say this because i like many of his songs but slow motion is definitely my literal least favorite. like, its just a really 'whatever' song out of most his stuff- composition and lyric wise. its not very intense. there are other songs i take far over that one more so due to the lyrics alone. i do feel like some stuff around that time was very "meh" actually and i have more so liked his recent stuff more than his older and sometimes more popular songs

 

[sketchesofpayne]

I'm not a huge fan of Pinocchio-P's music. Don't get me wrong, I'm not saying he's bad by any means, but his songs just feel very plain to me and lack something to really draw me in.

For me 70% of his music is skillfully executed but uninspired. The result of times when you want to create something but lack the spark of a great idea. So you make something that's... fine. I mean it's good but... eh. The guy's made, like, over a hundred songs since 2009. Not all of them are gonna stick the landing.

The other 30% is amazing and unique. There isn't much out there that's quite like it. My personal favorite is still "All I Need Are Things I Like." I could listen to that song all day long (especially the live Magical Mirai version and the Miku Expo 2021 version).

Some others I like: "Loveit" , "Because You're Here" , "Ultimate Senpai" , "The World Hasn't Even Started Yet" , "Nobody Makes Sense" , "Gorgeous Big Conversation" , "Apple Dot Com."

 

[peaches2217]

Gorgeous Big Conversation

Ah, I keep forgetting he was involved in that one! I mostly remember it for UtsuP's involvement. I also really like Rotten Heresy and Chocolate, thought I know it's suuuuuuuper super old.

 

[Purrpetuity]

My unpopular opinion is this: Myriad Virtual Singer is significantly better than both VOCALOID and UTAU

• Thirty built in voices, all computer generated, and all RealSinger voicebanks are free and available from Myriad's website
• Full voicebank creation with at most 100 recordings, full support for all 20 languages.
• Voicebanks are in the KILOBYTES - my largest one that's a RealSinger, supports every language and has icon data is 14KB
• Native support for 20 languages, including Romani, Latin, Russian, Chinese, Japanese, English (US and UK, with the right differentiations - they are not the same language), French and Romanian
• Recordings are single phoneme and easy to both modify and update
• You can submit your RealSinger voicebanks to be officially released
• Formant shifting as standard, allowing for more unique banks
• Word-based synthesis, unlike UTAU
• Full timed access for free, and the Virtual Singer software is only $15, with Melody Assistant being $30 (again, full timed access for free, so not necessary)
• Actively updated since it's creation in 1998, with the latest update being released in the past week
• Active beta tests where the creators actually listen to what people want
• Active forums where you can ask the two developers for advice and tips
• Active and close-knit community that isn't toxic in any way and is very helpful
• Built in tutorials, and both guides and manuals on the website
• You can use both MIDI and MusicXML

If I had to recommend anyone any synthesis software, it would absolutely be this

 

[MillyAqualine]

My unpopular opinion is this: Myriad Virtual Singer is significantly better than both VOCALOID and UTAU

• Thirty built in voices, all computer generated, and all RealSinger voicebanks are free and available from Myriad's website
• Full voicebank creation with at most 100 recordings, full support for all 20 languages.
• Voicebanks are in the KILOBYTES - my largest one that's a RealSinger, supports every language and has icon data is 14KB
• Native support for 20 languages, including Romani, Latin, Russian, Chinese, Japanese, English (US and UK, with the right differentiations - they are not the same language), French and Romanian
• Recordings are single phoneme and easy to both modify and update
• You can submit your RealSinger voicebanks to be officially released
• Formant shifting as standard, allowing for more unique banks
• Word-based synthesis, unlike UTAU
• Full timed access for free, and the Virtual Singer software is only $15, with Melody Assistant being $30 (again, full timed access for free, so not necessary)
• Actively updated since it's creation in 1998, with the latest update being released in the past week
• Active beta tests where the creators actually listen to what people want
• Active forums where you can ask the two developers for advice and tips
• Active and close-knit community that isn't toxic in any way and is very helpful
• Built in tutorials, and both guides and manuals on the website
• You can use both MIDI and MusicXML
  
  If I had to recommend anyone any synthesis software, it would absolutely be this

THIS. THIS. THIS. THANK YOU.

 

[Purrpetuity]

THIS. THIS. THIS. THANK YOU.

EXACTLY! And all anyone knows it for is Hello World, which didn't even use RealSinger! No one knows that it has that capability and it's such a waste because the engine is absolutely incredible for what it is!

 

[lIlI]

EXACTLY! And all anyone knows it for is Hello World, which didn't even use RealSinger! No one knows that it has that capability and it's such a waste because the engine is absolutely incredible for what it is!

Do you have any examples you can share? I've searched for Myriad Virtual Singer a few times on Youtube & Soundcloud and found very little, especially for English pop/rock.

 

[Purrpetuity]

Do you have any examples you can share? I've searched for Myriad Virtual Singer a few times on Youtube & Soundcloud and found very little, especially for English pop/rock.

It's built mostly for opera and choir stuff, but I did find an example of a cover of Meteor Sighting and Schwarze Regen

 

[MillyAqualine]

Moreover, what are great with making a RealSinger is that :

-It doesn't accept any recording until you reached a certain level of noise-free (or less BG noise) environment
-It accepts the phonem you record if you sing it well (as in, if it's too accented it won't accept it until you've made it correctly)

So now imagine if UTAU and even VOCALOId had these (though of course they don't act same as RealSinger so yeah)

Aaaaaand you can also switch languages in a same song if your DB covers more than one language by adding said language's code in a formula before the first lyrics you wanna enter using the other DB

And ! You can even have among the languages : Occitan (my area's language that is also a linguistic group that covers several Occitan variants from Languedocian Occitan -the one that I think is used in the program and that we used to speak and still somehow speak in my area- up to vivaro-alpin Occitan Valleys in close Northern Italia -and I don't remember it but it always seemed to me myriad's creators are based in Toulouse, hence why they added this regional language to the list), Latin (sadly not Greek but for opera and Gregorian stuffs it can be helpful), you have two distinct pronunciation in French ("Northern" being the standard spoken in places like Paris and "Southern" being close to my area's accents as well as fellow Southern places and regions like PACA) and even has both British English and American English (sadly no specific like Scottish or fellow big English groups like Australian)

So sure RealSinger may sound rustic (mind you it was released around or even prior to VOCALOID's first iteration) but regarding what Purrpetuity listed, it's still fascinating and worth a shot

 

[aru ii]

My unpopular opinion is this: Myriad Virtual Singer is significantly better than both VOCALOID and UTAU

• Thirty built in voices, all computer generated, and all RealSinger voicebanks are free and available from Myriad's website
• Full voicebank creation with at most 100 recordings, full support for all 20 languages.
• Voicebanks are in the KILOBYTES - my largest one that's a RealSinger, supports every language and has icon data is 14KB
• Native support for 20 languages, including Romani, Latin, Russian, Chinese, Japanese, English (US and UK, with the right differentiations - they are not the same language), French and Romanian
• Recordings are single phoneme and easy to both modify and update
• You can submit your RealSinger voicebanks to be officially released
• Formant shifting as standard, allowing for more unique banks
• Word-based synthesis, unlike UTAU
• Full timed access for free, and the Virtual Singer software is only $15, with Melody Assistant being $30 (again, full timed access for free, so not necessary)
• Actively updated since it's creation in 1998, with the latest update being released in the past week
• Active beta tests where the creators actually listen to what people want
• Active forums where you can ask the two developers for advice and tips
• Active and close-knit community that isn't toxic in any way and is very helpful
• Built in tutorials, and both guides and manuals on the website
• You can use both MIDI and MusicXML

If I had to recommend anyone any synthesis software, it would absolutely be this

Ohh, another software older than Vocaloid. I have heard about it 1 or 2 times, but I didn’t knew it was that great

 

[peaches2217]

Ohh, another software older than Vocaloid. I have heard about it 1 or 2 times, but I didn’t knew it was that great

Honestly? I think it's good, but I don't think it's great. It's definitely wonderful and gives an abundance of options - the more the merrier! But listening to it, I can't in good faith say I find it better than most Vocaloids and well-made Utaus. Still, it's cool, and I'll definitely be checking more of it out!

 

[Purrpetuity]

Ohh, another software older than Vocaloid. I have heard about it 1 or 2 times, but I didn’t knew it was that great

It's honestly really great and really cheap, though there is the setback that it isn't a piano roll, but sheet music. Harmony Assistant has a piano roll but it's $80 so there's that to keep in mind

Honestly? I think it's good, but I don't think it's great. It's definitely wonderful and gives an abundance of options - the more the merrier! But listening to it, I can't in good faith say I find it better than most Vocaloids and well-made Utaus. Still, it's cool, and I'll definitely be checking more of it out!

It is important to keep in mind that most samples use the robotic voices, and the RealSinger sample I sent was incomplete according to the creator. There are a lot of drawbacks on all sides but considering the price, age and active update status, I'd say those setbacks are well worth it.