Surprisingly, without me knowing, there is now a new version of AIVoice called AIVoice 2. The previous version of AIVoice is just an updated version of Voiceroid 2, which allows for expressiveness. However, I felt that AIVoice TTS quality sounds a bit rough compared to TTS software from other TTS software for Japanese like Cevio AI and VoicePeak.
Just several days ago, AIVoice 2 came out with some surprise. AIVoice 2 introduces new DNN voice synthesis technology to make the voice sound more natural. From testing it, it’s on par what VoicePeak provides, which is impressive on how good the sound quality is, which makes it suitable for generating listening content to practice with. Not to mention, there is finally a Mac version, although unlike Voicepeak, AIVoice 2 only works on Apple Silicon Macs.
At the present, there are four voice sets available, Kotonoha twins, Iori Yuzuru, Yuzuki Yukari and Kizuna Akari. Since I somehow own the AIVoice voice bank for Akari, here is a comparison:
AIVoice 1 Kizuna Akari (CV Yonezawa Madoka):
AIVoice 2 Kizuna Akari (CV Yonezawa Madoka):
At least there is another alternative for Mac users to use right now. For those who need to create listening materials from text (e.g. like news in Japanese), AIVoice 2 is now another viable alternative.
Currently, the voice sets for AIVoice 2 are 10% off on DLSite. Might be worth taking a look at if you are studying Japanese and need immersion practice or just want to create content in Japanese for output.
Leave a Reply