-Hitsune Kumi RVC-
Hitsune Kumi is created and voiced by Cubialpha (https://cubialpha.wixsite.com/kumivoice) and RVC models were created and trained by Haruki-P (https://utaforum.net/members/haruki.19900/)
Sample's uses the song 赤とんぼ (AKATONBO), UST by @mazapan_stuff (English translation made by me since i heard the first line and didn't know what was the full translation, so i made my own) and SOLARIA SynthV1 used for inference
-Info-
Hewwo :) (furry ahh spelling), thank you for downloading these databases :D
Hitsune Kumi RVC is a collection of 9 RVC models trained from Hitsune Kumi's UTAU voicebanks (each model uses the same name as their equivalent UTAU voicebank, except Normal and Legacy). These datasets and append info can be found here:
Normal: Basically her normal voice, uses her RVC dataset and engine samples from both her English and Japanese voicebanks
Dataset Length: 03:03:11
Quill: Softer/Quieter voice, uses engine and direct samples
Dataset Length: 01:14:45
Crayon: Cuter/Child-like voice, uses engine and direct samples from both her Redux and OG voicebanks (with a bias towards Redux)
Dataset Length: 02:06:30
Pigment: Powerful voice; can sound young at times, uses engine and direct samples
Dataset Length: 02:04:51
Erase: Screamo voice, uses engine and direct samples from both her Japanese and English voicebanks
Dataset Length: 02:06:22
Pastel: Weaker/Gentle voice, uses engine and direct samples
Dataset Length: 01:58:17
Lead: Chest-heavy power voice, uses engine and direct samples
Dataset Length: 01:12:25
Boyish: Masculine-leaning young voice, uses engine and direct samples
Dataset Length: 00:57:11
Legacy: Trained with all her older normal voicebanks, uses engine and direct samples from Act 3, Act 2, and Act 1
Dataset Length: 02:10:02
All models were trained using contentvec, rmvpe and HiFi-GAN; all models are also at 48000 Hz and have either 120 or 70 epochs
-Terms Of Service-
The ToS is the same as the ToS on Kumi's UTAU voicebanks, it can be found here: https://cubialpha.wixsite.com/kumivoice/character. There are some small differences and additions for when using these models: 1: Please credit Cubialpha as the voicer and the creator of Hitsune Kumi, and Haruki-P as the creator of the models
2: If you're going to use these models for normal speech, then don't make them say R-18/Sexual content (singing, short spoken segments and talkloids that include that type of content are allowed, but full speeches or similar content aren't allowed)
3: Please correctly acknowledge the use of these models if you use them in your content with the correct name. Correct naming examples include:
Hitsune Kumi RVC -Insert Append Name-
筆音クミ RVC -Insert Append Name-
Hitsune Kumi RVC (Insert Appends Used)
筆音クミ RVC (Insert Appends Used)
And that's all the extra rules, please consult her official ToS for her other rules :) https://cubialpha.wixsite.com/kumivoice/character
-Links-
Cubialpha: https://linktr.ee/cubialpha
https://cubialpha.wixsite.com/kumivoice
https://bsky.app/profile/cubialpha.bsky.social
https://soundcloud.com/cubialpha
https://www.youtube.com/@cubialpha
Haruki-P:
https://utaforum.net/members/haruki.19900/
https://soundcloud.com/ositostuff
Have Fun :D