WebKit

Introducing the Wheatley | Stephen Merchant RMVPE model, powered by RVC v2 and trained over 2500 epochs. This advanced AI voice model generates stunning AI music and covers, utilizing a massive 16 batch size and 30,000 steps. Leveraging 40k pretraining and incorporating vocals from popular games like Half Life 2, Portal 2, and Portal Stories Mel, this model delivers exceptional performance. With clearly labeled transposes in audio filenames, creating AI covers and text-to-speech has never been easier.

16 batch size RMVPE 7:43 minutes of audio ripped from Portal 2 30,000 steps 40k pretrain < *Infers were using vocals ripped from Half Life 2, Portal 2 and Portal Stories Mel* the transpose used are labeled in the audio file names

Wheatley | Stephen Merchant

Wheatley | Stephen Merchant RMVPE - RVC v2 2500 epochs

lusbert_

Portalfan

21 epochs on GPT 16 epochs on SoVITS both batch size 3 dataset size 17 minutes and 11 seconds 44kHz ripped from Portal 2

Cave Johnson - JK Simmons | GPT-SoVITS

20 epochs for both GPT and SoVITS batch size 4 for both as well dataset size of 23 Minutes i had 1 hours and 11 minutes but that'd be too large for me sooo also im surprised it did really good with the robotic voice, unlike RVC atleast, i didnt make an RVC model of her myself but all i've heard is that GLaDOS with RVC sucks. but well this one is actually robotic. most likely cause its over trained but it doesnt sound bad from that. so its a win win <

GLaDOS | Portal 2 e20-GPT_e20-SoVITS

TLDR the SoVITS is 15e with GPT being 16e GPT is DPO trained < notes: (for people who wanna learn more stuff about my experience with GPT-SoVITS) dataset size is 21 minutes batch size 2 on GPT and SoVITS SoVITS Learning Rate 0.4 15 epochs sovits took around 40 minutes 16 epochs gpt 27 minutes with DPO saving freq 2 at first i trained GPT to 20 epochs but last epoch saved is e18 prob cause my saving freq was 3 so ye be sure to look out ofr that as well as for SoVITS model it was fine most likely since i trained to 15 epochs and not 20 so prob smth to do with if the modulo of epoch is 0 or not so like [epoch count]mod[saving freq] with mod being % in python so in my case where i trained to 20 epochs with saving freq of 3 20%3 which is 2 aka non 0, so thats BAD :3 so just make sure the mod is always 0 so that you train to the full epoch count i didnt like the model when i first trained without DPO so i retrained GPT model with DPO it also went from using 6GB VRAM to using 8GB this time it took 35 minutes also apparently in SOME cases the default pretrained GPT model is better, often where new words are said or the punctuation is different i'd say? but im not really sure

TF2 Spy | Dennis Bateman_e16-GPT_e15-SoVITS

### this is a re-upload cause *SOMEONE "accidentally"* deleted the <#1175430844685484042> channel < Epochs: 1000 Steps: 15,000 RVC v2 Batch Size 8 Dataset size: 6:27 minutes all infers are transpose 0 since the Announcer has a robotic sound effect you should also apply effects in some DAW like FL Studio

Aperture Labs Announcer - Portal 2, 1000e, 15000s

Aperture Labs Announcer - Portal 2 | RMVPE - v2, 1000e, 15000s

### this is a re-upload cause *SOMEONE "accidentally"* deleted the <#1175430844685484042> channel < Epochs: 2500 Steps: 20,000 RVC v2 Batch Size 16 Dataset size: 3:44 minutes all infers are transpose 0

Wheatley | Stephen Merchant

Description

Samples

Saiba Momoi (Blue Archive)

Adel imam (Egyptian Actor) عادل امام (ممثل مصري)

JP Furina / Focalors CV: Inori Minase (Genshin Impact)

Ariana Grande AI

Hu Tao JP (Genshin Impact)

JENNIE of BLACKPINK [Strong Ver.]

Saiba Momoi (Blue Archive) (VA: Tokui Sora)

Hatsune Miku

SpongeBob SquarePants (Talking And Singing)

Takanashi Hoshino (from Blue Archive)

Satoru Gojo (JJK) [VA Yuichi Nakamura]

ENHYPEN Heeseung

Sunaokami Shiroko (Blue Archive)

Goku (Mario Castañeda) (Latin American Spanish) (UPDATE)

Villager (Minecraft)

Mortis [Brawl stars]

Jungkook (BTS)

Melodie - (Brawl Stars) - (English & Korean Voices)

Tendou Arisu (Blue Archive)

Kanye West