Koezu Yawa
UTAU Voicebanks
P / CV 4.0
Yawa's first voicebank, released in 2017.
Language: Japanese with minimal English support (based on paintedCZ phonetic system).
Encoding: CV / romaji only
P / 're:code' CV 4.5
A rerelease of Yawa's CV 4.0 bank with improvements to the oto.ini file made by Dyscha, leaving nearly 700 configurations rather than the previous ~330.
Language: Japanese with minimal English support (based on paintedCZ phonetic system).
Encoding: CV / romaji only
P / 'multipitch_strength' CV 5.0
Contains 'original' vocal (rerecording of Yawa at a natural tone and strength), one 'character' / falsetto vocal (harajuku), two 'pitch' vocals (masc and femme), and three 'strength' vocals (strong, gentle, and whisper). All vocals are accessible using the suffix broker. Oto.ini created on commission by Ryan-kun.
Language: Japanese with additional English consonants (L, R, and V)
Encoding: CV / romaji and hiragana
P / 'rainbow - original' CVV-VC v1.0
A lite version of Yawa's CVV-VC bank which only contains his 'original' vocals.
Language: Japanese with additional English consonants (L, R, rolled R, V)
Encoding: CV / romaji only
P / 'RGB'English VCCV v1.0
A single-pitch VCCV English bank using the paintedCZ system. Oto.ini created on commission by Ryan-kun
Language: English
Encoding: VCCV (paintedCZ phonetic system)
S / 'Cool' CVVC v1.0
A single-pitch version of Yawa's Multipitch CVVC bank, only containing the F2 pitch, deemed "cool". Provides a deep tone with average to soft strength. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and R, rolled R, and V.
Language: Japanese
Encoding: CVVC (5-mora, single pitch)
S / 'Emerald' CVVC v1.0
A single-pitch version of Yawa's Multipitch CVVC bank, only containing the A2 pitch, deemed "emerald". Provides a low to mid-range to soft tone with average to soft strength. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and R, rolled R, and V.
Language: Japanese
Encoding: CVVC (5-mora, single pitch)
S / 'Jade' CVVC v1.0
A single-pitch version of Yawa's Multipitch CVVC bank, only containing the F#3 pitch, deemed "jade". Provides a mid-range tone with average strength. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and R, rolled R, and V.
Language: Japanese
Encoding: CVVC (5-mora, single pitch)
S / 'Natural' CVVC v1.0
A bi-pitch version of Yawa's Multipitch CVVC bank, containing the Emerald and Jade banks. Provides a low to mid-range to soft tone with average to soft strength. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and R, rolled R, and V.
Language: Japanese
Encoding: CVVC (5-mora, A2 and F#3 pitch)
S / 'Neon' CVVC v1.0
A single-pitch version of Yawa's Multipitch CVVC bank, containing only the B4 falsetto pitch. Provides a unique sound type with a high, unstable growling tone and a hoarse, yelling strength. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and V; Lacking -y consonant blends.
Language: Japanese
Encoding: CVVC (5-mora, B4 pitch)
S / 'Blur' CV+
A single-pitch CV bank with additional "quick" samples. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and R, rolled R, and V, as well as CCV blends.
Language: Japanese
Encoding: CV(1-mora)
S / Multipitch CVVC v1.0
A multipitch CVVC bank, containing 10 pitches. Includes low to high range tones, along with vocalfry, growl, falsetto, and a variety of strengths. Coded in both hiragana and romaji. Contains additional English consonants of a hard J and R, rolled R, and V.
Language: Japanese
Encoding: CVVC (1-5-mora, F2, G#2, A2, C#3, F3, F#3, G#3, A#3, D4, and B4 pitch)
*CURRENTLY BEING OTO-ED AS OF 02/04/2024
Koezu Yawa
AI Voicebanks
RVC 'Classic' v1.0.e150
Yawa's first RVC voice model.
Made using voice samples from all of his voicebanks recorded from 2017-2024 (CV 4.0, CV 5.0, full 'Rainbow' voicebank including unreleased pitches -Original, Masc, Femme, Harajuku, Gentle, Whisper, and Strong-, RGB, Cool, Natural, Neon, and full Multipitch CVVC including Vivid, Emerald, Warm, Pastel, Jade, Cool, Dark, Saturation, Light).
*Includes the Dark and Neon data samples, which give some instability to the voice model, specifically on very low and very high notes.
Only trained to 150 epochs, higher training to be released in the future.
Sample Language(s): Japanese, English
RVC 'Classic' v1.5.e150
Alternate version of RVC 'Classic' v1.0.e150
Made using voice samples from all of his voicebanks recorded from 2017-2024 (CV 4.0, CV 5.0, full 'Rainbow' voicebank including unreleased pitches -Original, Masc, Femme, Harajuku, Gentle, Whisper, and Strong-, RGB, Cool, Natural, and full Multipitch CVVC including Vivid, Emerald, Warm, Pastel, Jade, Cool, Saturation, Light).
*Excludes the Dark and Neon data samples for model stability, specifically on very low and very high notes.
Only trained to 150 epochs, higher training to be released in the future.
Sample Language(s): Japanese, English
RVC 'LITE' v1.0
Made using voice samples recorded in 2023. Data comprised of approximately 2.5 hours of singing data.
Data Sample: ~2.5 hours
Sample Language(s): English, Japanese
*MODEL STILL TRAINING AS OF 02/04/2024