Vocalsynth Voicebank of the Month - December 2016

This month I'm gonna be talking about the Sharpkey voicebank Huan Xiao Yi, and as a bonus, will also be reviewing Sharpkey.

Huan Xiao Yi is a Mandarin Chinese voicebank for a new vocalsynthesis program named Sharpkey, developed by Chinese coder and producer Boxstar. Huan Xiao Yi was the first and, up until just the other day, only voicebank available for Sharpkey. She's powerful in her higher range, softer in her lower range, and is recorded in four pitches, with her optimal (in my opinion, anyways) range coming to around F#4~C2. She has a fairly feminine tone, the kind you hear somewhat often in other female vocalsynth voicebanks. Huan Xiao Yi works pretty well with faster songs, but she can't handle songs as fast as "The Disappearance of Hatsune Miku", because songs that fast are just a little bit too fast. Trust me, I tried covering that song in Chinese using her, and the fast part at the start sounded wrong no matter how much I fiddled with the parameters. Huan Xiao Yi supports note lyrics written in PinYin, and includes two dictionaries consisting of the most common pronunciations of a handful of the most common Hanzi characters, so you can sometimes type in Hanzi too, if you want to. This dictionary is also used with the other currently released Mandarin Chinese Sharpkey voicebank, Kiana, and will probably be used with every other Mandarin Chinese voicebank that gets released for Sharpkey.


Enough about Huan Xiao Yi, let's talk about Sharpkey itself.


Sharpkey's UI looks a lot like other software, like VOCALOID, CeVIO and MUTA, so users who're comfortable with those will feel right at home here. That is, if they know Chinese. Sharpkey is currently only available in Chinese, so for non-Chinese speakers, it will take some experimenting to actually figure out what does what, but it's fairly easy to figure out, and it's likely that once you know how to do things, you'll be fine.

Tuning in Sharpkey is also very easy, in fact, it's the easiest to tune out of ANY vocalsynthesis software I've used, and also lets you have the most control over the way things sound, unlike SOME software. (I'm looking at you, Alter/Ego.) While it is easier than tuning in VOCALOID, it's actually somewhat similar in how parameters work. You can actually draw pitchbends, but unlike VOCALOID where you draw in a little box underneath the pianoroll, you actually draw ON the pianoroll, so a lot less guesswork is involved and you actually know where the pitch is changing, how drastically it's changing and what it might sound like without having to fiddle with three parameters too much.

Pitch isn't the only parameter Sharpkey has, though. It also lets you control dynamics in a similar way you can pitch, how powerful or soft things sound, breathiness, one I don't actually know how to use and can't tell what it does, as well as two more which require more in-depth explanation. The two leftover paramaters are why I say that Sharpkey allows a lot of control over the way things sound.

One of them lets you change the duration of certain parts of samples, like the consonant length sometimes, the transition from the consonant to the vowel, the actual vowel, and the end of the sample where it either transitions to the consonant or vowel at the start of the next note, or where it slowly drops off and fades into nothing. The other reminds me of an unintentional feature of UTAU, where when a voicebank has multiple pitches recorded, you can append the pitch name to the end of the note so you can use a specific sample from a specific pitch if you want to. Sharpkey does this differently, where it displays the notes with a little speechbubble looking thing pointing to the note. The colour of the speech bubble pertains to the pitch source of the note. Warmer colours like red and orange are the lower pitches, and cooler colours like blue and green are the higher pitches. This feature may sound like nothing, but it can help if a voicebank, for example, is softer in it's higher ranges and more powerful in it's lower ranges. If you have notes in the lower range that you want to make sound soft, you can change it so Sharpkey grabs the sample from a higher pitch, effectively making the lower note sound soft instead. The last parameter there was just added in the latest update, which I didn't have while working on a cover of "Ave Maria" in Chinese to test Sharpkey, until it was too late and I had already posted the cover. This isn't bad, though, because it means I can compare what it sounds like unaltered, and with the samples grabbed from other pitches, something I'll attatch to the bottom of this post.


I have a lot of faith in Sharpkey, and if Boxstar actually plans on publicly releasing the voicebank creator like I've heard people say a few times, then I have a feeling it'll take UTAU's place as "that one free vocalsynthesis program everyone uses and makes voicebanks for". Or, at least I HOPE that's what happens. I'm also considering writing something about Sharpkey tuning tips in the future, maybe.


If you want to download Sharpkey and a voicebank or two to try out yourself, then you can head over to the official website if you have a Baidu account, or, check out this thread if you don't have a Baidu account.