Hoshikuzu Yuzuko (星屑ゆずこ | ほしくずゆずこ) (CHN: 星屑秋月 | Xìng Xiè Qiú Yuè) is a Japanese and Chinese vocalist for the DeepVocal engine. She is the first vocalist from Re:Voice.



Hoshikuzu Yuzuko debuted as a vocalist for the UTAU engine on April 5, 2017 although her character was conceived back in 2010. On January 1st, 2019, Mei-Saime made a post on the official Twitter of Re:Voice about name changes to the group and Yuzuko's surname, which became Hoshikuzu instead of Hoshine. Her new revised design was revealed on March 1st, 2019.

Since then, she has gained updates to her voice and even private voicebanks like Chinese and a recently released beta VCV styled bank on July 10th, 2019.


On July 14th, 2019, Mei-Saime made a post on Re:Voice's official Twitter announcing plans to make vocals for Yuzuko on the DeepVocal engine and plans to have Japanese and Chinese voicebanks released, with a Natural voicebank being teased on July 16th, 2019.

She was first released for closed beta July 27 - 28, 2019 with half of her vocal data in order to check for quality and errors, until days later in early August when she became available for open beta.

Yuzuko was released as a Technical Release version on September 29th with Mei-Saime stating that this release is not a full, "official" one as more extras and an additional pitch to be added to the Full release. She also stated that the 8th pitch, which is the falsetto pitch, was to be added at a later date due to bugs on the pitch.

On October 7th 2019, a short Mo Li Hua demo was made using an in developmental version of Yuzuko's Chinese voicebank for DeepVocal, but at the time, it was only CV and recorded with one completed pitch. However, on December 20th, a teaser UTAU CVVChinese demo was released and in the description, it stated that the DeepVocal version will contain the same amount of pitches and is in development as more Chinese voicebank related content is expected to be released in 2020.


A demo of her Chinese voicebank for the UTAU engine singing "风萤月" was posted on January 24th 2020 and it was confirmed that her DeepVocal Chinese voicebank was used in the background in the chorus, meaning that it was still in development. On January 26th, a full cover of Mitchie M's song "买买买" was posted featuring solely her DeepVocal Chinese VB along with a short demo of the song "玛德琳娜电塔 2.0". It was also around that time that Mei-Saime released Yuzuko's beta Chinese VB (consisting of 2 pitches) on the official QQ group for DeepVocal, for anyone to try out the build, with the desire to release more builds to improve her sound and pronunciation.

On the character's 3rd anniversary, her CHN voicebank was released with 3 pitches and in "build004", meaning she will likely be updated. Mei expressed that she will update her to have all pitches planned and that she released her CHN voicebank for people who would like to use her. It is unknown if her JP voicebank will also be updated, but it is also very likely, most likely a Version 3.


Yuzuko is a 17-year-old high school student and a mage who specializes in magic, and also psychokinetic abilities. Due to her "older-sister" type personality and the fact of her lineage heritage, she is given the codename: "Eternal Goddess". Her theme color is #FCAA95. As she goes into her new design, Mei wanted to portray a gentle, sister-like character, but brimming with beauty and an easy approach for newcomers to familiarize and to connect with. 

She is Re:Voice's first craft and a vocalist in Re:Voice's 5 girl series [Crystalia Millenium] as she is the "leader" and main character who specializes in magic and "pilots" the 0405 - HOSHI.


Her Japanese name, 星屑ゆずこ (ほしくずゆずこ), is comprised of the characters:

星 (ほし) - Star

屑 (くず) - Fragment

ゆずこ - Moon (According to what Mei envisioned)

Her name means, according to an interpretation of her theme by Mei-Saime is "Star Fragment of the Moon" in Japanese.

Her Chinese name on the other hand "星屑秋月" (Xìng Xiè Qiú Yuè) is comprised of the characters:

星 (Xìng) - Star

屑 (Xiè) - Fragment

秋 (Qiú) - Autumn

月 (Yuè) - Moon

Her surname, "星屑," means the same in Chinese as it does in Japanese. An appropriate English rendering of these characters would translate to "Star Fragment of the Autumn Moon".


Mei-Saime designed and did all of the illustrations of Yuzuko. As seen in the gallery, her design has drastically changed.

In her old UTAU art originally, her design had brown hair and blue eyes in an array of costumes in the lolita category from a maid styled uniform (2017) to a frilly type purple dress (2018).

Her current official has fair skin and long pink hair. She wears a white dress with a frilled collar with side frilled straps coming from each side. Underneath the white dress are light blue frills with mahogany tight and light blue heels. She also wears a black corset. On top of her head, she wears a yellow ribbon-like headband that holds up her bangs.



Demo 1 - "トゥール " (Tuuru | Tool) by MARETU YouTube bilibili SoundCloud
Demo 2 - "桜ノ雨 " (Sakura No Ame | Sakurano Rain) by halyosy YouTube bilibili SoundCloud
Closed Beta Release Cover - "蜜月アン・ドゥ・トロワ " (Honeymoon Un Deux Trois | Mitsugetsu An Du Torowa) by DATEKEN YouTube bilibili SoundCloud
Technical Release Cover- "Nebula " (Nebula) by Tripshots SoundCloud
Early Chinese VB Test- 茉莉花 (Mòlìhuā) bilibili
Chinese Voicebank Demonstration (Background Harmonies Only)- "风萤月 " (Fēng Yíng Yuè|Wind Fireflies Moon) by Zoey YouTube bilibili SoundCloud
Chinese Voicebank Demo 1- "买买买 " (Mai Mai Mai|Buy Buy Buy) by Mitchie M YouTube SoundCloud
Chinese Voicebank Demo 2- "玛 德 琳 娜 电 塔 2 . 0 " (Mardayline Tower) by 纯白 SoundCloud
Product Information
  Genre: Pop, Eurobeat, Jazz  Tempo: 90 - 220  Range: B3-E5
Basic Information

•Yuzuko Hoshikuzu's first UTAU voicebank was released back in April of 2017, with the Technical Release version of her DeepVocal voicebank being released on September 29, 2019. The DeepVocal voicebank has a mature, older-sister type feminine voice, using completely new recordings and not those based on her UTAU versions. The voicebank is more clearer and higher quality than previous UTAU incarnations.


• Yuzuko's Chinese voicebank for the DeepVocal engine. It is a natural styled voicebank, aimed to be smooth and of quality to her Chinese UTAU voicebank. It uses new recordings not found in her UTAU version.

Bank/Phonetic Information


• It was recorded for the Japanese language.

• The Technical Release version was recorded in 6 pitches, as the Full version will contain 7 pitches plus extras.


• It was recorded for the Mandarin Chinese language.

• Build004 has only 3 pitches, but the final version will have 6 to 7 pitches plus extras..

Software Issues as noted:


• Prior to the Technical Release, her main problem was that her vowel transitions were choppy. This issue has been fixed in the Technical Release.

• Some of her VCs do not register, but it is relatively smooth.

• She has some pops in her samples, but it can be reduced by tuning and mixing.


• Her voicer is an American with an American accent, so she has some pronunciation errors.

• Some of the "y" family Pinyin has VC errors, but can be worked around with tuning.

• T1 "han" has a distortion error, but can be worked around by switching Timbre to T2 or T3.





DeepVocal Voices
Community content is available under CC-BY-SA unless otherwise noted.