Kuroki Kasori AI

Kuroki Kasori AI for DiffSinger is a synthesized male Japanese voice, trained using artificial intelligence based on Shallow Diffusion Mechanism techniques. This voice bank is capable of naturally replicating the singing of the original voice provider, offering exceptional versatility through its four available vocal modes: Natural, Solid, Soft, and Falsetto.

With these modes, Kuroki Kasori AI can produce soft and tender tones as well as solid and aggressive ones, covering a wide vocal range that adapts to various musical and artistic needs.

Additionally, this voice model includes an Autopitch function that enhances the natural quality of its singing.

Vocal Modes

Natural:

This is Kuroki Kasori's default vocal mode, replicating his natural and distinctive singing. His voice is sweet and soft, with delicate tones that gain strength and power in the higher notes.

It's the most developed vocal mode of the voice model, with a total dataset of 1:43:11.

Recommended tension: 20 ~ 30.

Since this voice model has been trained with vocal modes the main voice became quite soft in tone vs. how the audios (wavs) are originally. I recommend using a tension of 30 to match the natural sound from their original recordings (as it sounded in version 1.3).

Diff Natural

00:00 / 00:18

Solid:

This vocal mode maximizes Kuroki Kasori's power, elevating his voice to a level of intensity that translates into strong and energetic shouts. It is ideal for performing aggressive songs in genres like rock, metal, and punk, providing the strength and dynamism necessary for these musical styles.

This vocal mode has a dataset of 44:36.

Recommended tension: 35 to 90.

Diff Solid

00:00 / 00:21

Data of the voice model

Acoustic: 152.000

Variance: 146.000

Dataset: 3:22:16

Natural: 1:43:11
Solid: 44:36
Soft: 43:54
Falsetto: 10:35

The voice model was trained with tension (vr) allowing to increase its strength or decrease it more easily and comfortably to any available vocal mode.

Special Phonemes:

AP/br/bre/breath: To activate the breaths

hh: To use stronger exhalations

cl: For short pauses between notes

Updates:

29/01/2024: Kuroki Kasori AI for Diffsinger 1.0
03/03/2024: Kuroki Kasori AI for Diffsinger 1.1
17/03/2024: Kuroki Kasori AI for Diffsinger 1.2
22/05/2024: Kuroki Kasori AI for Diffsinger 1.3
08/07/2024: Kuroki Kasori AI for Diffsinger 2.0

Kuroki Kasori AI 1.3 (old version containing only the main "Natural" vocal mode).

Kuroki Kasori AI

Vocal Modes

Natural:

Soft:

Solid:

Falsetto

Data of the voice model

DOWNLOAD

DOWNLOAD