Mafuyu Asahina (Project Sekai)
AnimeFictionalRVC v2Japanese

📣

From the creator

"I hope you like her! She was trained with ~15 minutes of singing so she's mainly suited for that. I want to make another model in the future with her talking voicelines. Also if the Baka Mitai sample sounds different than usual it's because I used Akiyama's cover not Kiryu's :)"

Description

Discover the power of cutting-edge artificial intelligence with our Mafuyu Asahina (Project Sekai) RVC v2 [Harvest] voice model. Trained over 465 epochs and 27,435 steps, this advanced AI model specializes in creating realistic and engaging singing performances, using as little as fifteen minutes of original audio data. Our Mafuyu Asahina model harnesses the potential of Retrieval-Based Voice Conversion (RVC), allowing it to transform text input into expressive speech or even generate stunning AI music covers. With its sophisticated algorithms, this next-level AI assistant delivers natural and nuanced output, making it perfect for various applications such as virtual assistants, entertainment platforms, and educational projects. Unlock unparalleled creative freedom by experimenting with different musical styles and genres. Generate captivating AI covers effortlessly or convert your favorite texts into life-like narrations. Transform your digital experiences with our versatile and efficient AI Voice Model—available now at Weights. Don't miss out on this opportunity; sign up for free access to our powerful suite of AI tools today and start exploring endless possibilities for your personal

See more

Weekly Metrics

Samples

New
Classic

Pitch

0

1

Male

Singing

🇺🇸 English

2

Female

Singing

🇺🇸 English

3

Female

Singing (Dry)

🇺🇸 English

4

Female

Singing (High)

🇺🇸 English

5

Male

Singing 2

🇺🇸 English

6

Male

Singing (Dry)

🇺🇸 English

7

Male

Singing (Dry, High)

🇺🇸 English

Comments

Sounds great!

why harvest??

I didn't know about rmvpe at the time. I'm currently remaking the model with a higher quality dataset and will be using rmvpe :)

update i feel like the new one sounds worse T_T now i'll try the old dataset with rmvpe

Add a comment

Post