There is a new, kind of mysterious web AI text to speech called CoeFont STUDIO. (It's a weird spelling of koe (voice).)
Website: クリエイターのための無料で使えるAI音声合成サービス

The homepage lets you try Allial's (ARIARU) voice with a text box, but you can't try out Millial (MIRIARU) without logging in because she's a prototype. You can easily log in with a Google Account.


They are made by YELLSTON and were established in November 12th, 2020. CoeFont STUDIO became public today.


v flower enthusiast
Jul 14, 2018
honestly quite like the designs, but im also just happy to see some pure black/white designs because id love to see em more lol. tested out the voices a little to and i think they sound really good, a lot less engine noise than i would expect, even from a prototype. theyre also pretty smooth, even when i threw in some english phrases it wasnt too bad

i have next to no use for tts so i dont expect to buy anything but the two voice choices are pretty good and im curious to see how it develops!
The only products they have is CoeFont STUDIO and something called "The very difficult AI test (超難問AIテスト)" where you have to guess if the voice is a human girl's or an AI.
The test asks you to figure out if she's a human or not and to type your name in. I got 2/5 right and you basically can't hear a difference. The character's name is Erst (エルスト / ERUSTO).

Strangely, the homepage for YELLSTON itself has a weird mission statement that includes the word "erst".

The Japanese says "Togatta gijutsu de WAKU-WAKU o souzou-suru" which means "Create excitement with cutting edge technology".
"Erst" means "first" (as in "earliest") in German, I am very confused.

The company page says
Yellston combined the German word "erst" to mean "advanced/superior" and the English word "yell" to mean "cheer for".

It also says they're aiming on making their company have cutting-edge AI tech.


kiyoteru enthusiast
Staff member
Apr 8, 2018
I got 4/5 right on the quiz and 5/5 on the second try (it uses different audio each time) - for the most part I could pick up the AI quirks (some words sounded a little choppy or "doubled" but I had to take a second listen just to make sure I heard it), but it is really impressive.

It sounds really nice! Minimal noise and the results are really smooth. Definitely interested in seeing how it progresses.
  • Like
Reactions: morrysillusion

Users Who Are Viewing This Thread (Users: 0, Guests: 0)