Originally uploaded on September 3, 2023
Image link: https://i.kym-cdn.com/entries/icons/facebook/000/038/462/Untitled_2.jpg
Model link: https://huggingface.co/joebamnasoda69/CheckardsArrVeeSeeStuff/blob/main/RalseiV4.zip
Original description:
Probably the closest thing we'll get to Ralsei having an actual canon voice.
Dataset created by taking his little dialogue noise, stretching it out, EQ-ing it through heck and back to make vowel sounds, splicing it with some of the in-game sfx to make consonants, making a model out of it, generating some pretty decent audio with said model, cherry-picking the audio I generated and adding that audio to the dataset, and making a model out of that dataset to slightly improve pronunciation. (Neither Elevenlabs nor TortoiseTTS were used.)
Trained by me, no credit needed.
NOTES: Pronunciation will still sound a bit off at times, so I highly recommend either setting the index feature ratio to a lower value (<0.55), OR using a totally different index file altogether. Also, I recommend turning the protect consonants value to somewhere around 0.15.

Additional Details:

Tags: Fictional Character, RVC V2

Related URL: https://i.kym-cdn.com/entries/icons/facebook/000/038/462/Untitled_2.jpg

Related URL: https://huggingface.co/joebamnasoda69/CheckardsArrVeeSeeStuff/blob/main/RalseiV4.zip

Download Link: https://huggingface.co/joebamnasoda69/CheckardsArrVeeSeeStuff/blob/main/RalseiV4.zip