Other StudioVoICE Discussion Thread

[VocAddict]

Official Website​

This is a thread for the discussion of StudioVoICE, a new AI-based synthesizer developed by Microsoft STCA. It is a cloud-based vocal synthesize software based on their on AI System, Xiaoice. You can download it for free on their website.

Voices available (with raw tests):

• Xiaobing (Xiaoice)

• He Chang (Luna)

• Chen Ziyu (Alpha)

Software Translation by @cocoacacao

 

[VocAddict]

It's nearly midnight and I'm too tired to dissect all of this information . My knowledge of Chinese is 0.5 so I've been glossing Google translate and textbooks for a while. But Xiaoice. Microsoft's AI System that does a lot of stuff apparently and its new newest venture is singing. Here are some demos, example of the software.

On the page itself, there are sample buttons you can hear the current singers in Chinese. There's most likely a lot more but I haven't done a deep dive into it yet. In the footer it says 'Statement: The relevant content of this product is generated by artificial intelligence technology' which is pretty neat.

One thing to note about the software itself, it is cloud-based, so basically Emvoice in nature. But these guys have far larger ranges, though they are still restricted in the same manner of Emvoice. You need to log in with an account to export as an audio file (it appears to be Chinese mobile devices only?) but you can record it through desktop audio in Audacity or some other DAW. There are three dependencies that you need to install before you can run it, and it's listed in a text file for you to easily get them if you don't have it installed. You may have to run the program as administrator for the first time to get it to run.

Xiaoice Japan (Rinna) - Avex


LUNA (Chinese)

https://www.bilibili.com/video/BV1qa4y1v7im/?spm_id_from=333.788.videocard.1

 

[VocAddict]

In regards to the software itself in terms of usage, it's really nice! I haven't experienced any bugs yet (except some instances where it wouldn't render and I was unable to figure out what the error meant) and in their raw states, they're pretty clear! There isn't any parameters to tune at the moment, and it appears that it only allows one vocal track and one instrumental per project. You just input your notes and lyrics (pinyin is not supported as far as I can tell), choose whether you want initial or ending breaths (you can add them manually or let the software decide for you) and click play. It's 158 MB when extracted.

Here's a picture of the interface in use to get an idea of how it looks.


It comes with in-built filters to give the voice a more "full" sounds (it kind of sounds like stereo-wide reverb being applied honestly) but you can export it raw without any filters being applied. There's volume/gain and panning as well. And as you can see, it supports the trusty [-] for extending notes.

Other than that, there isn't much more (I still have to go through the menus themselves but I doubt I'll find anything new that the manuals through machine translation hasn't shown me. It comes with two projects and two manual files (which I've attached so someone can actually translate it lol). When you start the software for the first time, it has a short tutorial of where things are, which is nice.

Ranges of voices:

• Xiaoice: G2 - G5
• LUNA: B2 - B5
• Chen Zi Yu (Alpha): C2 - G4

 

[VocAddict]

Profile images because that might be useful as well.

 

[haru0l]

I was wondering if Microsoft will ever release their synth to the public and here we are!
First impressions of this are very good imo

 

[peaches2217]

I'm loving the way it looks and sounds! It just finished downloading, so I'm gonna install it now and try for myself!

 

[peaches2217]

So I've been tinkering around with it and with the PDF files for the past hour or so, and I've gotta say, I like it! I wish there was a way to edit the pitch bends (maybe in the future...? Neutrino has an editor, this could too!), but other than that I like the way it sounds in-editor!

(I did Luna dirty. I really should have lowered the key for her... Luna baby I'll make it up to you)

​

The PDF files basically just cover setup, the purpose of the vocal flavor thingies, making tracks, exporting, and keyboard shortcuts - nothing groundbreaking or tricky to grasp. There's also Ventilation and Pause, which I need to mess with more... the PDFs kinda gloss over them, so I can't say too much about them yet, but they seem to function similarly to Vocaloid's Velocity parameter. It's also noted that you can't adjust the position of any accompaniment you import, so you'll have to adjust it before importing it or just build your midi around it.

 

[VocAddict]

I wish there was a way to edit the pitch bends

Yep, the lack of a pitch editor is really depressing but I have my hopes that it can improve since it is still in beta and Microsoft is accepting feature and ideas (smiley face in the title bar).

The PDF files basically just cover setup, the purpose of the vocal flavor thingies, making tracks, exporting, and keyboard shortcuts - nothing groundbreaking or tricky to grasp. There's also Ventilation and Pause, which I need to mess with more... the PDFs kinda gloss over them, so I can't say too much about them yet, but they seem to function similarly to Vocaloid's Velocity parameter. It's also noted that you can't adjust the position of any accompaniment you import, so you'll have to adjust it before importing it or just build your midi around it.

Ah, so I basically figured out the software by clicking around lol. I guess it was as simple to use as I thought it would be and maybe that was what they wanted when they released this synth.

​

I believe the first 2 options refer to the ventilation/pause parameters they were speaking of. I'm still having issues rendering (keep getting "server is busy, please try again later") so it's a pain so I'm just going off what I heard the first few times it did work and extrapolate from there. There is an option in the fifth dropdown menu, first option that seems to deal with the automatic application of parameters?

The first option makes a 'V', which puts breathes wherever you apply it which is nice for the beginning of phrases, as well as in between long phrases to give the appearance of breathing lol. The second option makes an 'O' tag, which I'm guessing is for pauses mid-phrase? Perhaps glottal stops or just stops instead of smooth singing?

The third option is just the normal note.

The one thing is the choice between setting lyrics from this note forward, or setting lyrics for the whole track (the fourth and fifth respectively). There isn't an option to add lyrics for a selection of notes but that can change as well if they deem it necessary.

 

[VocAddict]

Also, I was somehow blind and did not notice that their names are on the illustration.

Blue boy name is Alpha.

 

[VocAddict]

I rendered the iconic Chinese bank test and popular folk song, Mo Li Hua to give those an idea how how they sound in their raw state without anything changed.

Xiaobing (Xiaoice)


Chen Ziyu (Alpha)


He Chang (Luna)

 

[nalondollie]

This is really cool! I was wondering if something like this could come about when I first learned about Xiaoice and am happy to see it come true! And not only did we get her voice but multiple ones! I love the tone with Luna's voicebank.

Overal, I'm really excited to see how this software progresses! Thanks for the hard work too, @VocAddict !

 

[VocAddict]

Rinna is based on the same framework that the trio use but for Japanese in partnership with Avex. Here's her latest song.


And she performed at DIVE XR FESTIVAL, a collaboration event where digital performs Vocaloids to Virtual YouTubers and even gacha bands perform.

 

[VocAddict]

This is really cool! I was wondering if something like this could come about when I first learned about Xiaoice and am happy to see it come true! And not only did we get her voice but multiple ones! I love the tone with Luna's voicebank.

Overal, I'm really excited to see how this software progresses! Thanks for the hard work too, @VocAddict !

Thank you and you're welcome! I'm really excited to see how this progresses.

On their site, there is this silhouette so we should keep an eye on her to see when she is released. I might be making matter out of nothingness here but I wonder if she's an English voice because of how the other image files are name.

Xiaoice - xiaoice_bg.png
Luna - hechang_bg.png
Alpha - chenziyu_bg.png
Yellow No Name - singer_uk.png

​

 

[cocoaCacao]

Xiaoice's name in pinyin is Xiao Bing (bing means ice), Luna's is He Chang, and the blue boy is Chen Ziyu ("Chen" being a traditional Chinese surname, "zi" meaning child, and "yu" meaning jade), I think alpha is a placeholder name denoting his status as a WIP voicebank.

One of the prompts the singer can give you is "I prefer Chinese songs. With more training I could sing in English, etc." which implies the possibility of English and Japanese voicebanks in the future.

Here's a translation guide I've made for the software:

 

[lIlI]

Rinna has one of my favourite voice types in Japanese synths, almost entirely due to her expressiveness.

 

[VocAddict]

Xiaoice's name in pinyin is Xiao Bing (bing means ice), Luna's is He Chang, and the blue boy is Chen Ziyu ("Chen" being a traditional Chinese surname, "zi" meaning child, and "yu" meaning jade), I think alpha is a placeholder name denoting his status as a WIP voicebank.

Here's a translation guide I've made for the software:

Thank you for the translation!

Yeah, I was aware of the hanzi for bing meant ice and it's most likely the reason why it's highlighted the way it is in VoICE, should have mentioned that too lol. And the alpha thing makes sense but honestly at this point, the only thing I can call him is blue boy and alpha lmao.

One of the prompts the singer can give you is "I prefer Chinese songs. With more training I could sing in English, etc." which implies the possibility of English and Japanese voicebanks in the future.

Ah, that's cool. I was wondering what they were trying to tell me the entire time lol. Thanks again!

 

[cocoaCacao]

Ah, that's cool. I was wondering what they were trying to tell me the entire time lol. Thanks again!

They mostly just yell at you for notes that are too short, tell you that they can only sing in Chinese, and say that changing the length of a note can change the pronunciation, it's kinda funny, and np!

 

[PigeonP]

I've just used all three with vocalshifter, they're very impressive!! Transcribing and entering lyrics was very intuitive. The way the lyric edit system will keep the notes ahead intact and let you enter new notes worked sooo nicely, reminded me of cevio. Since my reading in Chinese is limited it really saved my ass, I still made a mistake in Nature, noticed last minute. :c

Despite the lack of parameters I find myself really enjoying using it, the tone is just so nice. Vocalshifter fits like a glove! Luna had the most noise using it. I don't think I've ever worked this fast haha.

I can't choose a favorite. Perhaps Chen Zi Yu, but I love some of the cute tones from Xiaoice and I love the falsetto of Luna.

Here's I LOVE U with Xiaoice

Here's ACE with Luna

Here's Nature with Chen Zi Yu (forgive my instrumental, couldn't find any good acoustic guitars to do it with)

I can't wait to see where StudioVoICE goes in the future! I'l certainly be doting on it the entire way! :P

 

[VocAddict]

We might be able to register with international numbers in the near future if things go well.

 

[VocAddict]

In regards to restrictions and protocols of the software this is their response. Tl;dr, we can use it for personal works but not commercial.