Overview
KAFU (可不, Kafu) is the flagship vocal synthesis library of KAMITSUBAKI STUDIO’s “Musical Isotope” project, powered by CeVIO AI and Synthesizer V, with voice provider KAF.
Released on July 7, 2021, KAFU became a phenomenon in VOCALOID producer culture, featuring in viral hits like Tsumiki’s Phony.
VOICE PROVIDER
KAF
SINGING ENGINES
CeVIO AI; later expanded to Synthesizer V AI
CHARACTER DESIGN
RELEASE
7 July 2021
Design and voice identity
PALOW.’s character design uses white and yellow, circular pupils, a hooded outer layer, and geometric details. It makes KAFU legible beside KAF while establishing a separate voice-character identity for creators.
Timeline
| Period | Milestone | Notes |
|---|---|---|
| October 2020 | Voice survey | Three prototype singing types were presented; Type B became the basis of the released library. |
| November–December 2020 | Early covers | Hinadori, Shinzō to Karakuri, and Kako o Kurau were published as early demonstrations. |
| May–July 2021 | Commission series and release | A broad producer lineup introduced the library; the final release arrived on 7 July. |
| Later | Engine expansion | The product line grew beyond its CeVIO AI launch with a Synthesizer V AI edition. |
Commissioned-song index
Show the launch-era song index
| Date | Song | Creator |
|---|---|---|
| January 9, 2021 | Hologram | Hifumi |
| February 20, 2021 | Cute na Kanojo | syudou |
| March 7, 2021 | Ice Cream | Guiano |
| April 10, 2021 | Retropolis | R Sound Design |
| May 1, 2021 | Hana to Nare | Yunosuke |
| May 8, 2021 | Suimenka | Osanuma Paseri |
| June 5, 2021 | Phony | Tsumiki |
| July 17, 2021 | Cat Loving | Moimi Kashii |
These official commissions are an entry point rather than a closed discography: KAFU’s significance lies in the expanding body of songs made by independent creators. Phony, Marshall Maximizer, and Cute na Kanojo are widely cited examples of this ecosystem.
Representative Works and Related Entries
Expand full singles list
Singles
| Release date | Title |
|---|---|
| 2010-12-12 | 不可欠 |
| 2020-02-29 | 病名を教えて |
| 2020-12-24 | くるくる |
| 2022-01-29 | メイドインベイビー |
| 2022-05-11 | 夜の自販機 |
| 2022-07-12 | 水面下 (feat. Kafu) |
| 2022-07-12 | 流線形メーデー |
| 2022-07-31 | ようやく君が死んだんだ。 |
| 2022-11-09 | ハロウィン・マスカレイド |
| 2023-02-28 | イーシャンテン・オーバードーズ |
| 2023-03-17 | 生涯、貴方を悔やみます。 |
| 2023-03-23 | ストレイシープ |
| 2023-04-29 | ルムルトル |
| 2023-05-03 | アノニマス |
| 2023-07-21 | スワロー (feat. KAFU) |
| 2023-12-10 | 解/体 |
| 2024-01-08 | おさけを飲んだらくすりが飲めないの唄 (feat. 初音ミク) |
| 2024-02-26 | 模範解答実験室 |
| 2024-04-04 | Voyager |
| 2024-05-31 | 混沌ブギ (Cover) |
| 2024-09-05 | 最期の逢瀬 |
| 2024-10-15 | コぇちっちゃ<てゴ×ンネ |
| 2024-10-22 | 『抱きしめて。』 |
| 2024-12-20 | アット・ザ・電脳シティ(STEAKA Remix) |
| 2025-02-01 | 人間に産まれたから |
| 2025-02-20 | MaDiVA |
| 2025-04-18 | テレパシ (feat. 重音テト) [Cover] |
| 2025-07-11 | Plazma (feat. 重音テト) [TETO with KAFU Cover] |
| 2025-10-05 | IRIS OUT (feat. 重音テト) [Teto Cover] |
| 2025-10-13 | IRIS OUT (KAFU Cover) |
| 2025-10-26 | IRIS OUT (feat. 初音ミク) [Hatsune Miku Cover] |
| 2025-12-17 | モニタリング (feat. 重音テト) [Cover] |
| 2026-05-21 | 痣痣 |
| 2026-07-28 | ゆらゆら |
Show the launch-era song index
| Date | Song | Creator |
|---|---|---|
| January 9, 2021 | Hologram | Hifumi |
| February 20, 2021 | Cute na Kanojo | syudou |
| March 7, 2021 | Ice Cream | Guiano |
| April 10, 2021 | Retropolis | R Sound Design |
| May 1, 2021 | Hana to Nare | Yunosuke |
| May 8, 2021 | Suimenka | Osanuma Paseri |
| June 5, 2021 | Phony | Tsumiki |
| July 17, 2021 | Cat Loving | Moimi Kashii |
Official Compilations and Naming System
Built from KAF’s voice, it is the first entry in the Musical Isotope series and by far the most widely propagated, with the deepest influence on the Chinese-language and global creator communities.
| Item | Detail |
|---|---|
| First compilation | KAF+YOU KAFU COMPILATION ALBUM Symmetry (2021-11-24) |
| Naming | Follows the formula “source-witch code + YOU”, meaning “KAF and you” |
| Voice template | KAF |
| Release system | KAMITSUBAKI STUDIO Musical Isotope series |
What the naming means: titling the official compilations “X+YOU” redefines the Isotope from a “tool” into “a partner who creates alongside the listener”. Every track on the compilations comes from an invited or selected creator, making them a direct intersection of the official and community creative ecosystems.
Acoustic Engines and Library Specifications
The Musical Isotope project is built on a dual-engine acoustic architecture, each engine serving a different purpose.
| Engine | Developer | Technical character |
|---|---|---|
| CeVIO AI | Techno-Speech | Evolved from a hybrid of HMM and deep neural networks (DNN), it captures vibrato, breath and vocal-fold micro-noise — the “imperfect but alive” detail of a real singer — to an extreme degree |
| Synthesizer V AI | Dreamtonics (5th generation) | A fully autoregressive acoustic model with cross-lingual singing (natural Japanese/English/Chinese transitions), seamless morphing of emotion parameters (Power / Soft / Clear), and very high render speed and editability |
Why both: CeVIO AI preserves human imperfection; Synthesizer V AI handles cross-lingual work and deep editability. Together they cover everything from reproducing a live vocal texture to giving global creators full creative latitude.
Open Licensing and the Derivative-Work Ecosystem
For the Musical Isotope project KAMITSUBAKI abandoned the closed restrictions of conventional artist management and built an open, self-reinforcing licensing loop.
- A commercial-use threshold instead of a ban: doujin musicians may freely use the voice libraries within a defined revenue ceiling and release the resulting originals straight to streaming platforms.
- Official compilations that absorb the scene: through open calls for official compilation albums, outstanding community creators and works are brought onto major releases, creating a cycle of “community creation → official validation → wider reach”.
- Results: this mechanism produced phenomenon-level UGC hits such as “キュートなカノジョ” and “フォニイ”, carrying AI singing into the centre of mainstream pop culture.
Industry significance: where the traditional record system restricts derivative works, KAMITSUBAKI treats a voice library as a gateway into an ecosystem rather than a closed asset — the fundamental reason the Isotope family accumulated such a vast body of UGC within a few years.
Landmark UGC Hit Archive
Once released, the KAFU library triggered a phenomenon-level explosion across the Vocaloid and doujin scenes. The following tracks are landmarks of that wave.
| Track | Creator | Significance |
|---|---|---|
| ”キュートなカノジョ“ | syudou | Fused KAFU’s breathy timbre with a sickly-sweet groove, becoming one of the fuses for her surge in popularity |
| ”フォニイ“ | Tsumiki | Its slashed drums and falsetto turns made it one of the most played and most covered songs in KAFU’s history, sung by countless artists at home and abroad |
| ”マーシャル・マキシマイザー“ | Hiiragi Magnetite | Showcases the extremes of KAFU’s mechanical texture and agility through ultra-dense beats and rapid-fire diction |
Historical position: these works appeared at the turn of the Heisei and Reiwa eras and are widely regarded as the turning point at which AI singing entered the centre of mainstream pop culture. KAFU is consequently the best known and most cross-community Isotope in the KAMITSUBAKI family.
Project connections
KAFU opened the Musical Isotope project. The subsequent SEKAI, RIME, COKO, and HARU voice characters extend the same creator-facing framework.
Relationship to the Source Witch
| Item | Detail |
|---|---|
| Voice template | KAF |
| System | One node of the five-witch isotope matrix of V.W.P |
| First release | CeVIO AI (a Synthesizer V AI edition followed) |
| Position in the lore | The digital-side counterpart of KAMITSUBAKI’s “song singularity”, sharing one source of “singing voice” with KAF |
Naming rule: each Isotope’s official name is drawn directly from its source witch (KAFU←KAF, RIME←RIM, SEKAI←ISEKAIJOUCHO, COKO←KOKO, HARU←HARUSARUHI). The convention itself declares the core idea: different forms of one existence.
