GUMI is now a Voiceroid too!

Megpoid Talk is a new – and, to me at least, completely unexpected – product that gives the singer we more commonly know as GUMI proper speaking abilities via a separate voicebank and user interface.

So far at least, it seems that it is only in Japanese; but as Megpoid English already exists as a Vocaloid, it would be no surprise if an English version of Megpoid Talk were to appear as well, either with this or in the future. I have no idea on this point as yet, but shall keep watching developments.

The new package will be released in February 2014 and will contain a range of expressive and other controls, a selection of what might be termed 'mood sounds', and even allows one to speak into the software and it will have Megpoid speak it in the same tone and inflections – this last feature being somewhat like VocaListener. It can also handle vocaloid-format sound files (VSQs).

I do not know what its intended applications are, but no doubt the style and location of advertising of the new product will reveal distinct clues in this regard.

The following short demo gives a very good idea of what will be possible, and it seems to do it mostly perfectly. Time will tell whether there are any significant issues, but my initial impressions are very favourable....

New Zunko Model

Zunko Tohoku now has a MikuMikuDance (MMD) model, which is really nice and is demonstrated well in this clip. This is apparently a 'school uniform' model, though this outfit isn't quite like any of the usual Japanese designs, of which there are three main types (the sailor-style, the maroon string-ribbon tie, and the neck-height bow of various styles).

The new model can be downloaded from a link about two-thirds of the way down this page. I am tempted, but have managed to resist the desire to download it myself. There are also a number of 'normal' and 'chibi' illustrations in two sizes each that are also accessible from that page. I now have a few of those for my library.

Zunko is a Voiceroid, so is like a Vocaloid but speaks instead of singing. In practice, it is possible (with some effort) to make either emulate the other's capability, and several producers have done just that in the past, so we can that it can be done.

In this case, it is so good, sound-wise, that it could even be just a human speaker and the character animated to fit the voice (as, I believe, is done with Synthesized Reality Productions' Yume) but there are a few moments in this that suggest it could well be genuine. One has to listen very carefully, and it is subtle...

Obviously, this is just a lot of Japanese, so few visitors here will be able to understand what Zunko is saying; nevertheless it is interesting to note her range of vocal inflections and to watch the MMD model in action...