Luka - Honkai: Star Rail
TTS / RealtimeFictionalRVC v2English

πŸ“£

From the creator

"(CV: Howard Wang) 40K Trained on 16 minutes of in-game dialogue"

Description

Introducing Luka [EN], the advanced AI Voice Model powered by Retrieval-Based Voice Conversion (RVC) technology. Based on the beloved character from Honkai: Star Rail, this model has been trained on 40,00

See more

Weekly Metrics

Samples

New
Classic

Pitch

0

1

Male

Singing

πŸ‡ΊπŸ‡Έ English

2

Female

Singing

πŸ‡ΊπŸ‡Έ English

3

Female

Singing (Dry)

πŸ‡ΊπŸ‡Έ English

4

Female

Singing (High)

πŸ‡ΊπŸ‡Έ English

5

Male

Singing 2

πŸ‡ΊπŸ‡Έ English

6

Male

Singing (Dry)

πŸ‡ΊπŸ‡Έ English

7

Male

Singing (Dry, High)

πŸ‡ΊπŸ‡Έ English

Comments

7.6 k step with 200e that mean dataset up to 1 hour

not rlly no

my 150 epoch model with 1 hour and 40 minutes of audio has 39.3k steps

mybe its take 8 hour

nope

image.png

not this one either

image.png

whay your pc spec

2080 ti 9900k

OK Explains a lot But how can 16 min be? 7.6k step

dataset is 16 mins

batch size 8

strange i use 30 min and batch 8 on 150 ep just 8-7 k step why

Β―_(ツ)_/Β―

Add a comment

Post

Privacy

β€’

Terms

β€’

DMCA