Hareford DiffSinger
This is a primarily English DiffSinger voicebank, and can only be used in OpenUTAU. There is data included to support Japanese, Mandarin Chinese, Korean, and Spanish. It is suggested that this voicebank is used between E2~E5.
This version of the vocal is in beta, and is trained on 37 minutes of Hareford's singing data, and 20 minutes of extra data.
All solfège's presented have minor equalization done, with increased volume.
Default
The default voice color will give Hareford's natural performance. This voice color is trained with 18 minutes of data.
Hareford DiffSinger Solfège
Power
The voice color "Power" will give Hareford's brightest performance. It is recorded with high energy, retaining his strength up to around C5. This voice color is trained with 2.5 minutes of data. The opposite of this voice color is "Soft".
Hareford DiffSinger Power Solfège
Soft
The voice color "Soft" will give a soft performance. It is recorded to target softer songs while avoiding sounding breathy. This voice color is trained with 3.5 minutes of data. The opposite of this voice color is "Power".
Hareford DiffSinger Soft Solfège
Solid
The voice color "Solid" will give a solid performance. It is recorded with a consistently solid tone, and gives Hareford a mature tone. This voice color is trained with 1 minute of data. The complement of this voice color is "Dark", and the opposite is "Cute".
Hareford DiffSinger Solid Solfège
Dark
The voice color "Dark" will give Hareford's softest performance. It is recorded with a low, breathy, and mature tone, giving softer results than the voice color "Soft". This voice color is trained with 30 seconds of data. The complement of this voice color is "Solid".
Hareford DiffSinger Dark Solfège
Cute
The voice color "Cute" will give a cute performance. It is recorded with an even squeakier tone than Hareford's normal tone. This voice color is trained with 30 seconds of data. The opposite of this voice color is "Solid".
Hareford DiffSinger Cute Solfège
Shout
The voice color "Shout" is recorded with shouting data. It slightly matures Hareford's tone, and gives him more power than his "Power" voice color. This voice color is trained with 1 minute of data. This voice color is experimental.
Hareford DiffSinger Shout Solfège
Blue
The voice color "Blue" is recorded to work specifically with jazz. It gives Hareford a mature tone, and also gives his silibant consonants a slight lisp. This voice color is trained with 5 minutes of data.
Hareford DiffSinger Blue Solfège
Rhythmic
The voice color "Rhythmic" is recorded to work specifically with rap. This voice color is trained with 1 minute of data. This voice color is experimental.
Hareford DiffSinger Rhythmic Solfège
Extend
The voice color "Extend" is trained with Japanese, Mandarin Chinese, Korean, and Spanish data. Using this voice color will improve pronunciation when using languages besides English.
Hareford DiffSinger Extend Solfège
Hareford DiffSinger Demonstrations