WebKit

Introducing the ericdoa RVC V2 Model, a state-of-the-art AI voice generator powered by Rico's VoiceConverter technology. This advanced model was trained over 400 epochs using the Mangio-Crepe 9-minute dataset, resulting in highly realistic and expressive text-to-speech capabilities. The ericdoa RVC V2 can even generate impressive AI music covers, making it a versatile tool for musicians and creators alike. Experience the future of AI voices today - try creating your own AI covers or text-to-speech with our free RVC V2 demo.

ericdoa

threshold_

i could just fantasize 😵‍💫 
trained with mangio-crepe, 9 minute dataset, 9200 steps
(the audio samples linked on this message are old, updated samples are at the bottom)

https://huggingface.co/tThreshold/ericdoa/resolve/main/ericdoa.zip

it actually is, but im currently training a new version of it with more data

this is the third ericdoa model i've made (i haven't made a model in like 4 months so feedback would be nice) trained on a 13 minute dataset, which consists of the whole doa album (mangio-crepe, 80 hop length) credit me if you want, but it'll be nice if you did

ericdoa (DOA Album)

ericdoa (DOA Album) (RVC V2) 300 Epochs

this is one thing i can do while i wait for dancinwitsomebawdy to drop trained on an ~11 minute dataset, consisting mostly from the 'things with wings' album (mangio-crepe, 64 hop length, 13050 steps) what makes this model different from the first version is that i trained this on ACTUAL .WAV AUDIO, my first version used .mp3s pulled from youtube credit me if you want, but it'll be nice if you did

ericdoa v.2

dataset is literally just eating chips, looped for SEVEN MINUTES credit me if you want, but ill appreciate it if you did

literally just eating chips

literally just eating chips (RVC V2) 300 Epochs (voicemy.ai)

to my knowledge, the other english gumball models posted here were trained on only singing (i think), so i made one trained on both speaking and singing (mostly speaking) trained on a 5 minute dataset from the episodes The Grades, The Diet, & The Faith (rmvpe, 4200 steps) voice actor: Nicolas Cantu transpose works best when between 6-12

Gumball Watterson (The Amazing World of Gumball)

Gumball Watterson (The Amazing World of Gumball) (RVC V2) 300 Epochs

i originally posted this model on kits, then i retrained it on rvc (it sucked) and then i reretrained it and it now sounds better trained on a ~6 minute dataset (rmvpe, probably 2000 steps?) credit is optional, but it would be nice since i would know where my ai model is being used :) (samples have backing vocals since i use covergen_no_ui for inferring)

ericdoa

Description

Samples

ericdoa v.2

ericdoa (DOA Album)

ericdoa (Low Taper Fade Guy)