#replivoice

Adachi Rei is agender, transmasc and pansexual!

tomo

adachi (utau)

Tomo Kane is a japanese synth released in 2011, and devloped by SHINYUSHA for the windows 100% magazine (more about that on 13 jul post). utau wikidot says tomo was released on 13 instead of 12 though. she is voiced by Tomoko Kaneda, and illustrated by 今泉昭彦(Imaizumi Akihiko ? maybe ?). tomos official mmd model was created by Aun.N. tomo comes with the windows 100 utau starter pack and the utau master pack. shes available in the nov 2011 issue of windows 100, and then received an update for the nov 2012 issue.

adachi rei is a japanese synth created and illustrated by missile39. missile39 created 2 life-size robots of hatsune miku in the past, and is now working on building a robot of his original character adachi rei. this is being crowd funded to pay for the material needed in 3d printing her different parts. adachis utau vb wa released on 12 oct 2019, and her A.I.VOICE (or more specifically REPLIVOICE) (developed by Mechanical Girl LLC) bank on 23 oct 2021, after a crowd funding campaign. adachis voice is completely computer generated using audacity through overlapping Sin waves and the use of equalizers to adjust the wave frequencies to match those of japanese speech (?). as such, she does not have a vp. a fan-made port to deepvocal was created by M@RS, and a conversion to vcv by K.C.. adachi is 160cm tall.

There’s some exciting UTAU related news today! Adachi Rei’s campaign to get a commercial text to speech voicebank has opened on Campfire.  As of this post, it’s 11 percent of the way towards its 3,500,000 yen/32,225 dollar goal. The A.I.VOICE stretch goal will be reached at 30,000 yen/300 USD. The rest of the crowdfunding will be for the building of an actual Adachi Rei robot.

Context

A.I.VOICE is made by AI, Inc, the same company who developed AITalk. Products like Voiceroid, Gynoid Talk, and Megpoid Talk all use AITalk as a base. AI, Inc also created Iori Yuzuru and Kotonoha Aoi/Akane, three successful Voiceroids. A.I.VOICE uses the newest version of AITalk, which was finished in 2020.

An UTAU being ported to A.I.VOICE would be a big deal, on the same level as Macne Nana becoming a Vocaloid. But Adachi Rei also isn’t like any other UTAU, and her talking voice likely won’t be a straight port. Why is that?

The UTAU without a voice provider

Adachi Rei/ 足立レイ was created by missile_39 in 2019. Her voice is synthesized not from recordings of a person, but by overlapping sine waves in Audacity to create many different types of sounds. Missile refers to this method of synthesis as “Replivoice” (Replica voice), which is what her product name will be if the crowdfunding succeeds. Her voice sounds a little similar to other “synthetic” UTAUs like Utane Uta, but she is in a class of her own by being a completely synthetic voice. She is an example of people hearing a voice from a random sound, a kind of auditory illusion.

There is an explanation video of how she was made here, but it’s long and untranslated.

Usually, text to speech voices are made by recording sentences and parsing words from a human source. But Rei has no voice provider, so there isn’t anything to record. The creator remarked that Rei sounded very similar to her UTAU version when a prototype was put into A.I.VOICE. It’s interesting to know that an established company is willing to promote something different like this.

Rei in real life?

About the robot stuff, Rei was actually designed with the intent of being a real-life robot before she was made an UTAU. missile_39 is a mechanical engineering student who has made a life-size Miku robot before. Rei’s body will be 3d printed out of resin, but the material cost is expensive, which is why it’s part of this crowdfunding campaign. Missile also wants to make robot Rei able to walk. 

Other rewards include a keychain of Rei, an MMD model of Rei with a unique number on the jacket for every supporter, and a special shout-out in a later announcement video.

If the campaign exceeds its goal amount, Rei will get additional emotion databases (happy, sad, etc) which can be toggled between.