Character Driven TTS Software from Z Vocal Project!
Hello everyone, if you’ve checked our website at all in the past month you might already be aware this is out, but if not, we released our new text to speech software “Z Idol Talker” on November 9th 2023! The program was made in about a week just as a test, and we decided to release it to the public for the hell of it. We are working on a couple more small features to be released soon, as well as tutorials, but don’t expect anything crazy to come out of this software unless there’s overwhelming demand. Feel free to report any bugs you encounter through this google form.
What is Z Idol Talker?
Z Idol Talker is a free standalone Unity application for using Piper Models. Voicebank management is made easy, with you being able to see a list of all voicebanks in your folder, alongside their name, language, voice type, and an icon to represent them. In addition, each voicebank has it’s own banner to represent itself once active (We are working on allowing a _0 image to serve as a talk sprite). In addition, each voicebank is able to load it’s own custom assets to transform the UI to match it’s theme. If you don’t want to fully customize the UI for your voicebank you can also just set a color for the default UI to change to.
A default theme for you to edit can be found on the main Itch.io page.
What languages does Z Idol Talker support?
Z Idol Talker supports the following Languages: Arabic, Catalan, Czech, Danish, German, Greek, English, Spanish, Finnish, French, Hungarian, Icelandic, Italian, Georgian, Kazakh, Luxembourgish, Nepali, Dutch, Norwegian, Polish, Portuguese, Romanian, Russian, Serbian, Swedish, Swahili, Turkish, Ukrainian, Vietnamese, Chinese
Unfortunately, Piper doesn’t have Japanese support so it is not fully supported in Z Idol Talker. The good news is that you CAN get it working in a kind of hacky way by training a Japanese voicebank as a Spanish voicebank and either using romaji inputs or changing the espeak and language codes to “ja” in the model’s json file, allowing hiragana inputs.
We are also attempting to get a base Japanese model working, but currently timing is completely messed up, and it might be an issue with espeak itself. (If you think you can get it working, the language code trick should work on the pretrained model, )
What first party voicebanks has Z Vocal Project released?
As of 11/24/2023 we have released the following voicebanks:
Z Vocal Project Characters: Dunder Announcer, Dunder Castilian, Dunder Spanish (Latin), Dunder French, ㄬomeder, and Ascenia
Voicebanks trained by Z Vocal Project: Defoko English, Yuto, Dario, and Tsukuyomi-Chan English.
Yuto and Dario were trained on CC0 audio, Tsukuyomi-Chan English and Defoko English were trained on UTAU samples. Defoko English may get removed as we do not have appropriate permissions.
ZVP Character banks and Defoko English can be downloaded on the application’s main page. The rest can be found on this page using the password “ZIdolTalker”. Demos of all these voicebanks can be viewed in the Club Room.
Wait who the hell is ㄬomeder?
ㄬomeder is the newest Z Vocal Project character, being a gnome that somewhat looks like Dunder despite the pair having never interacted with each other. ㄬomeder has an extremely nasally voice and in addition to his talk bank, has a CVVC UTAU in the works.
What updates are coming in the future?
Mainly we are working on the ability for the voicebank banners to allow for a speech sprite so that it looks like they’re talking alongside the audio that’s playing. It shouldn’t take that long I just have other things to do right now which is why this article took so long to make. Better UI Scaling is also being worked on.
Another thing we’re looking into is a built-in streamer mode, allowing the software to read out stream alerts and putting the character in front of a greenscreen, which will probably take longer for me to figure out.
Additionally as mentioned prior, I do want to get Japanese working, but I’m completely out of my element in this regard.
We are also working on text/video resources, but ended up being busier than expected this month, sorry about that.
What was changed in v1.0.1β?
v1.0.1β (Released on November 10th) fixed a bug that was causing only the first sentence of each line to play and an issue where accessibility mode and export location toggles were displayed wrong.
Thanks for reading!
As per last article, we’re still working on getting Dunder_CV out before the year’s end, however Fuuka S!ckBeta might be canned at this point. See you guys next month maybe?