SynthV Synthesizer V Discussion Thread

[Rylitah]

Synthesizer V's Christmas update has been announced, and it's... something to allow JP voicebanks sing in Chinese/English, and vice versa for English/Chinese banks? I'm not really sure of the details or if this just means it'll automatically convert phonemes to the respective languages, but it sounds really interesting.

The update will drop at around 6 PM Japan time (about 6 hours from the time of this post). Also, it's for Synth V Pro only.

 

[Silver1063]

Hmmmm, I really hope it isn't just custom dictionaries. I've seen research that translates a speakers voice into another language that they can't speak using AI, and its very convincing. Hopefully its something like that.

 

[Rylitah]

There's an article that goes into it with some example videos. It does seem to be deep learning - there's a sidebar menu that has a dropdown list that you can click to have the voicebank sing in the selected language. It says you can just enter the lyrics of the language and then set the option to that language, which seems simple enough.

The accent of the singer will be kept. Maki is mentioned as her JP and Eng banks still being useful since you'll get different results from her English bank singing in English and her Japanese bank being made to sing in English this way, as it will sound like a "Japanese-accented English voice" <- direct quote from the article.

Examples with Eleanor and Qingsu both singing in Japanese:

 

[Exemplar]

Official demo from Dreamtonics

 

[Silver1063]

I'm completely flabbergasted.

 

[REDD]

AAAAAAAAAAAAAAAAAAAA IM SCREAMING THIS IS REVOLUTIONARY

This is a HUGE step forward!!! This broadens our horizons for voicebanks so much! The language barrier? Who is she? Don't know her!!
oH man
like
could you IMAGINE if we had this feature YEARS AGO? with Vocaloid? Sure the voices are slightly accented, but it's WAAY more legible than say, the V3/V4 Crypton English vocal banks. AI technology has opened up ALL NEW FRONTIERS for us!!! Voicebanks are no longer limited by language! Now people in different language markets can use the same voicebank without having to wait for language updates!!!
I'm so excited guys you have no idea

 

[Blue Of Mind]

SynthV has become OP now, and I'm scared lmao.

More seriously, I hope this means it will be less of a hassle to implement future languages into SynthV, since technically you won't need to record separate language banks anymore. Okay, I just really want to hear German and Russian results with this feature.

 

[lIlI]

This is absolutely awesome! Previously most of my favourite voice types were in languages I didn't speak, now my wallet will never be safe again. I love Qingsu's voice, but I'm not skilled enough to manipulate her phonemes for English - she sounds perfect here! This update essentially trebles the potential of every voicebank.

*edit*

Canqiong English sample!

She sounds like a native speaker.

 

[Exemplar]

Just imagine what could happen next with the software. You realize how much money a voicebank maker could make if synthv can also do this in korean, spanish and french??? If Dreamtonics went with that idea of adding those 3 languages on top of the current 3, it's possible that over the next two years we could get as many banks as the Vocaloid 3 days... made at a fraction of the cost of a voicebank maker doing business with Yamaha to produce them.

If everyone can get their stuff together we could be entering a singing synth renaissance that can possibly cross into the mainstream. I think folks in the community might be underestimating what this actually means. The profit margin of what would have been considered a niche voicebank on Vocaloid back in the day could be significantly greater than a bank as popular as Miku or Gumi.

I'm not joking at all. A ton of cash spent to produce a single Vocaloid voicebank in a single language vs a moderate amount of cash spent to produce a SynthV bank in 3 to 6 and possibly many more languages in the future. That also means more money on hand for a company to promote their products with concerts and such.

Yamaha are going to be kicking themselves after realizing this could have been avoided if they kept Kanru around and Crypton & Internet needs to be giving him a phonecall ASAP.

 

[lIlI]

So here's an interesting test: a video comparing Maki English singing English and Maki Japanese singing English using the update:


They sound shockingly close, I don't think we'll be needing many natively bilingual voicebanks in future. Although it isn't 100% perfect yet, the technology will only improve. I'd be curious to hear how Maki Japanese sounds compared to Maki English singing Japanese.

*edit*

Here's another sample of a voicebank that wasn't shown in the demo: Tsuina-chan English

 

[pico]

Qing Su synthesized to English sounds ridiculously clean and Rikka is so cute. Like everyone else, I'm completely blown away. I hear people's concerns regarding SynthV being harder to manipulate for expression than other synths, but in terms of, like, raw quality and general advancement in this field, nobody has anything on SynthesizerV.

They uploaded another video featuring tsuina-chan, if anyone wants to see the feature using her library:

I think the final effect of this feature will hinge somewhat on the voice library itself and how it sounds, but "accented" english banks have been pretty much the standard for so many years-- it's no wonder that Maki AI ENG and Maki AI JP don't sound too different at all, despite Maki ENG being purpose-built. I have no complaints whatsoever. This feature is just so powerful and could create so many new creative avenues for producers.

I wish we could hear more beloved synth vocals on this engine. I love all of the new vocals and characters but I can't stop myself from wondering how my old favorites would fare with SynthV's engine and features.

 

[Nezuh]

Holy fuck, this is so awesome!
It's so perfect it seems fake. And I know it's not, I mean it as a compliment.
This is so revolutionary, there's no competition anymore. Synth V is the best in its class.


A Christmas miracle, you could call it. (lol)
---
Side note: I think I remember seeing a prototype that could do something like this years ago?, mind you it was for other software (either TTS Google or maybe Voctro) but still.

 

[morrysillusion]

woke up to check out these demos and wow wow wow. as someone who religiously makes japanese (namely flower) voices sing english this is just insane to hear. makes me wish all my favorite japanese voices would go to Synth V just for this. and it really broadens the horizon of what i can use. there were multiple japanese/chinese voices i liked the tone of, but never considered buying because of the effort to making them sing english- sure it could be done, but with no real support for that it obviously had some troubles. as much as i appreciate the full english vbs, the seamlessness between the voice and language due to this feature is really cool too- as too many times with vocaloid the switch between the english bank and og language can be a bit jarring. theres so much in store for this development and i am more than excited to see it. even if this isnt "perfect" for all languages, this is certainly a development no one has really tried before and its far far better than i wouldve expected.

 

[Buck]

I remember having a discussion about how getting rid of 1-language only voicebanks was something that would be a huge deal and something vocal synths should be actively striving toward

I think it has really been a problem in the past for a lot of people that the characters they liked couldn't really be used outside of a language they don't speak or (more importantly) don't listen to or write music in. A lot of people have their fav, but if their fav doesn't have a voicebank to sing in the language they actually intend to make music or covers in, they either get to deal with it sounding bad, or they can compromise and get a different voicebank for a character that they don't care about.

I get that recording 80 different voicebanks for each language isn't economically responsible for the developers, but I was still kind of surprised that in the year 2021, 17(?) years after vocaloid 1 was released, nobody had thought of a way to make multi-lingual singers more efficiently. But it seems like SynthV have figured something out with their AI tech, so I'm really glad this is finally at least 80% a reality (they just have to get more languages now like spanish, korean etc.)

So now if the reason why you like SynthV is because you are a fan of Medium5, you don't have to choose between their mediocre english capability or a native english voicebank anymore, that is, whenever they get updated to AI.

Speak of the devil:

 

[blackout]

Wow this is legitimately incredible. My biggest gripe with SV AI so far is the lack of diversity when it comes to voices especially on the JP side but this update just expanded the roster of all its supported languages with how this affects ALL AI banks both old and new. I'm so much more excited for all future releases of whichever native language now!!

The results are really impressive to me too, like I really can't tell the difference when they switch from Maki JP to Maki ENG in the english singing part. Saki and Qing Su's are also just... Chef's kiss on all three languages. Can't wait to mess around with this once I get on my PC.

 

[kozet]

Now if only we could get cross-language singing within a track. Imagine having English’s [θ] and [ɹ], Chinese’s separate aspirated and unaspirated stops, and Japanese’s vowels all at once. The only thing I could desire on top of that would be a good [ɬ] or [ʕ].

 

[poser]

this makes me wish vflower, sachiko or other unique vocalo voices were on synthv even more T_______T

 

[Blue Of Mind]

this makes me wish vflower, sachiko or other unique vocalo voices were on synthv even more T_______T

I wouldn't be surprised if more VBs on other synths jumped to SynthV in the future. It feels like its technological capabilities are rapidly improving with every update now, and with the cross-language support for all AI banks, it's now leaps and bounds ahead of Vocaloid, CeVIO, AIVoice, etc.

 

[aru ii]

Something is coming in 2 days

 

[aru ii]

The SynthV 1.6.0 update is here with 2 new major features

Synthesizer V Studio 1.6.0 Update | Dreamtonics株式会社

With the 1.6.0 update, Synthesizer V Studio Pro now supports switching between a few predefined vocal styles for AI voice databases. Vocal Mode Creators can now use the "Vocal Mode" section

dreamtonics.com