[SynthV Update 1.5.0] "Cross-lingual Singing Synthesis"

Wowwwwwww. I wanna know how much work went into the phonemes on that! Lots of nitpicky editing, or is cross-language that good? Either way it’s impressive.

The whole point of cross-lingual synthesis is that any AI voice can “understand” phonemes for the selected language without any nitpicking. While I’m sure there was some manual adjustment, it’s very likely that it was no more work than it would be for a native English voice like Anri or Eleanor.

1 Like

O that’s so good its unreal. WOW

I’ve got the money for Cangqiong set aside, so as soon as the cross-lingual version is available, I’m getting it. I just finished up a song using Anri, and I think Cangqiong would have been a better fit.


Hopefully ANiCUTE will be carrying it. :crossed_fingers:

1 Like

I also want Cangqiong, she’s amazing. Do you know when the AI version is coming out and where to buy it? I can’t seem to find that 😵‍💫😵‍💫.

Nope, see my prior posts. It looks like ANiCUTE is carrying the Standard version, not the AI version.

1 Like

I’d also recently heard a really nicely tuned GENBU voice. So I had hopes that maybe it would be available as a cross-lingual voice.

But based on what I read, it doesn’t look like there’ll be an AI version of GENBU in the immediate future.

I thought I had read somewhere that AI versions of voices could be trained from Standard voices, but I guess even if that’s true, there are additional licensing issues involved.

I guess we’ll just have to be patient with this. Things will be ready when they are ready. :expressionless:

still waiting for the english male voice to be released :frowning: … Early 2022 is right now in my book !

1 Like

any insider info on when it will land ? :slight_smile:

Considering there was no news at all for 6 months following the initial teaser, I wouldn’t expect any more news until a release date is decided. Keep in mind that for business purposes “early 2022” could be any time between now and the end of Q2.

1 Like

yes argh. I’m still hoping for a release very soon. I thought they might release it at the same time as Solaria, but that was just a sweet dream (Solaria being supposed to drop tomorrow). Imagine ? Both male and female native english AI the same day ? Would be the biggest step for the english users.

It’s entirely possible they don’t want to make people choose between the two by releasing very close together. If they announced a release date right now they might cause Solaria to get fewer pre-orders. It is generally unwise business practice to release too many products too close together.

1 Like

yes sure :slight_smile: ! Personaly I’ll happily buy both ^^’.

This is a great feature! Currently it seems like you can only use one language per track, would it be possible to switch languages mid-phrase in the future as is common in stuff like Japanese City Pop?

(Maybe this is already possible, but I can’t see a way to switch language in the same track)

Is Kyomachi Seika AI not capable of using Cross-lingual Synthesis?? I just bought her and currently she has no options available for any language other than Japanese, maybe because she only just came out??

1 Like

Are you using the Pro edition of SynthV Studio?

Yep! I have Solaria and Ellie and it works fine for both of them but not Seika. The only thing I can think of is if she were somehow locked to Japanese; I can’t find anything stating that she is or isn’t capable of using AI, but it would seem strange to me to lock her out of a feature like that. It is still the 26th in the US so maybe she’s not fully usable until tomorrow??

Currently this is not supported. I’m sure it’s not impossible for Dreamtonics to implement, but there would be a couple big obstacles:

  1. The phoneme lists for each language have overlapping symbols. For example, there needs to be some way to differentiate between English r and Japanese r. I suppose you could select language per-note instead of per track, but that brings me to the second obstacle:
  2. A major part of vocal synthesis is producing not just each phoneme, but the transition from one phoneme to another. When you increase the size of the phoneme list, the number of required transitions increases significantly. It’s one thing to make an AI voice sing in a different language, it’s a whole other thing for it to be able to seamlessly transition between languages.

Seika has two voice libraries: one Standard, and one AI (as well as lite versions of each). As long as the voice selected for the track says “AI” in it, you’re using the AI voice.

As for why you don’t have cross-lingual options, I’m afraid I don’t know. The only requirements should be that you be using a paid AI voice in SynthV Studio Pro. I guess Dreamtonics never specifically said that all future AI voices would support cross-lingual synthesis, but it would be odd for that to actually be the case since it doesn’t seem to be any extra work to implement for each voice.

It is possible that she’ll receive an update within the release period to add cross-lingual functionality, so perhaps keep an eye on the “License and Updates” panel.

This was my exact same thought! Given how Cross-lingual synthesis works, it shouldn’t be unavailable for any voicebank with AI capabilities but they never did explicitly say " it will be available to every future AI bank" and Seika has no promotional material acknowledging the feature, maybe just because they want to use it in a demo still before releasing it to the public?

To me, it would then seem unfair to charge the same for Seika as other AI banks who can use said feature if it is the case, but I’ve already spent the money so it’s a moot point. I’ll make sure to post an update in the next few days when she does/if she doesn’t update so others can be aware of the situation. Thank you for your help!

1 Like