Today’s update has introduced cross-lingual synthesis for all AI voices. This feature is exclusive to SynthV Studio Pro, and you must update both the editor and the voice library to access the new functionality. The demo also includes the unnamed English male AI vocal that was teased in the gen3 video last June.
Quadimension has also announced that they will be “updating” their existing voices to leverage this functionality. It is somewhat unclear what form this will take since the existing versions are all Standard libraries, but it seems they’ve created AI libraries based on the existing Standard ones. As of yet I haven’t identified how users will acquire the new versions since the change in format likely means it is not as simple as updating the existing installation.
Edit 1: Adding link to official release notes. This includes an announcement that Dreamtonics’ English male AI voice will release in “early 2022.”
Wouldn’t this make it much easier to get a single voice to sing in a conlang, cuz more phonemes (phonemes from Japanese, English, and Mandarin) are available?
Yeah… someone on Twitter did some side-by-side comparisons. As someone who also has both EN and JP Maki, while there are still some differences, I’m not sure it justifies double the pricetag any more…
This is amazing, one of the best features implemented in the software. Now people can enjoy every voicebank, because there’s no longer a language barrier. I can finally use the Mandarin voicebanks I thought I’d never use. The problem I have now is deciding which AI do I get first lol.
While the announcement says that “all” AI voices support the cross-lingual singing synthesis, it appears that this doesn’t include the Lite versions.
Or maybe this hasn’t been rolled out for Lite voices yet?
For example, I have Tsurumaki Maki AI ENG Lite, and it shows as v101. The version that supports cross-lingual is 103, and there are no updates available for the Lite voice.
Similarly, Eleanor Forte AI Lite v100 only supports English, and shows no available updates.
On the positive side, there are a few other additions:
Synthesizer V Engine: cross-lingual singing synthesis for AI voices (only for Pro edition), configurable via “Sing in the following language” in Voice panel.
Piano roll: automatically shrink the neighboring note when enlarging a note; togglable via “overlap-less note resizing” in Settings panel.
Other: add crash recovery.
The note resizing is really helpful, so I’m glad to see it added.
They said this will not be available for the Lite AI voices. Only the paid AI voicebanks will have this feature. Don’t know if this is temporary or not.
While I can’t say for certain, given that this is a feature exclusive to the Pro version of the editor I would be very surprised to see it extend to lite voices.
Especially since lite voices are primarily intended to serve as a trial or demo for the full product and the voice’s default/native language makes for the best representation of that.
It’s understandable, but sort of discouraging that the top announcement says “any Synthesizer V AI voices can sing in Japanese, Chinese and English.” without any sort of qualification.
The exception is buried at the bottom, in the release notes.
In fact, I missed seeing it even though I had pasted them into my own post!
But if it’s not “any”, that’s what it should say up top, and that’s not what they chose to write.
Edit: I had originally written “all”, but the announcement says “any”. I’ve changed my post to reflect the language in the announcement.
Access the Licenses and Updates panel on the right side of the editor. Click “Check for Updates”, then select the product you wish to update and click “Update”.
From what I’ve read, there’s been some miscommunications between Quadimension and Dreamtonics, so between retraining, encoding, and manually contacting the current users, it’ll take some time for existing users to get the AI versions of the Quadimension voices.
I didn’t realize the Quadimension voices Cangqiong, Chiyu, Haiyi, and Shian were available on the ANiCUTE page.
I’m not sure if AI versions of these voices will be available as separate products. It’s also not clear if all owners of the the Standard voices will get AI versions. If that was the case, I’d be tempted to purchase one now just to get in early.
Of those voices, Cangqiong is the one that’s the most interesting to me as a cross-lingual voice. She’s been described as a more “mature” voice, and from what I’ve heard seems to have more “power” than Eleanor Forte or ANRI.
She’s apparently the only one who’s currently been encoded and sent to beta testers, which is why there are some demo songs posted:
I haven’t heard enough demos to know for sure, but I’m actually a bit more excited on working with Cangqiong than Solaria, so… good job, Dreamtonics.
That means I’ll be waiting (im)patiently along with everyone else for more demos and for the AI versions to become available.
From my experience Cangqiong is the most flexible of the Medium5 voices. She has the most mature tone and performs well at a variety of ranges.
Shian is rather soft (perhaps most comparable to Kaai Yuki), whereas Haiyi and Chiyu both have a more striking, powerful tone.
That all said, it really does depend on how much the repackaging process changes or maintains each of their traits. I’m definitely looking forward to more demos as well.
