📁 dataset: 2 min 🤵 Creator: <@701828747568873583> 🔗 Download: https://huggingface.co/dragoncrack/https___www_donationalerts_com_r_crack_dragon/resolve/main/TobeyMaguire.zip 💸 Donate: ||https://www.donationalerts.com/r/crack_dragon 🥺|| ⚠️данная модель защищена от перезаливов!без разрешения запрещено выкладывать на любые другие источники/сайты/файл обменики без спроса создателя модели⚠️ ⚠️this model is protected from overloading!without permission, it is forbidden to upload to any other sources / sites / file changes without asking for a model⚠️ ♥️ write orders for the model in the dm, do not wait for me to add friends, write immediately🐬 ♥️ пишите заказы на модели в лс, не ждите, пока я добавлю в друзья, пишите сразу🐬 🎖️ also you can dm me or my friend <@597420942594932756> to request a model🎖️ 🎖️также вы можете написать мне или моему другу <@597420942594932756> чтобы запросить модель🎖️ https://m.media-amazon.com/images/M/MV5BMTYwMTI5NTM2OF5BMl5BanBnXkFtZTcwODk3MDQ2Mg@@._V1_FMjpg_UX1000_.jpg

https://youtu.be/4e4_9RdU2qI

Why only 2 min dataset?

why more?

You know

wouldn't it be better quality?

the quality of the dataset is better than the quantity of the dataset

no, the length of the dataset doesn't guarantee you quality, it's the quality of the dataset that matters on the quality of the model

yes but you can gather more than 2 minutes of the same quality

It's not important to make a good quality model

if you can gather more data of the same quality or better, then increasing the dataset will reduce artifacts and voicecracks and whatnot

No it won't, usually increasing dataset is required just to add more variety to the voice, like whispering, or singing, but it has a high chance of increasing artifacts, not reducing them

by this logic, just straight up taking my datasets and chopping them in half, and using like say the first half of the data, will create a better model

because if i had a 2 minute dataset vs me only training based on the first minute of data, then the 1 minute model would have less artifacting

It might help if your dataset had artifacts and you got rid of parts with artifact. But if your dataset was fairly same through the whole length, then it won't make that much of a difference quality wise, but it will increase the speed of training, which is also helpful

fair enough

i mean if the data youre adding is high quality then its better to add it then not

unless you have time concerns

adding bad data could lead to artifacting, but adding good data would reduce it

If you found a data that's better and at least longer than a minute and it has the same manner of speech then it's better to use that and scrap what you had before, because now the previous data that was worse might ruin your model with artifacts. I found myself this formula: 2-3 minutes of a good quality speaking, 1 minute of singing, if you want the singing voice of the person. That's more than enough to make a good model, but sometimes even 30 seconds is enough to make a good model

https://discord.com/channels/1159260121998827560/1163145701866348554/1163145701866348554 like here, I used 30 seconds of clean voice and it even managed to save the manner of speech of that person

fair but your original point is that adding more data increases artifacting

obviously you need the highest quality data

if you have a 2 minute dataset of hq data, and get 3 more, its better to train with the 5 rather than the 2

Yeah, but again, if you're not professional enough, you might add more unnecessary noise that rvc won't like. I'm not thinking about rvc being imperfect, I'm thinking about people being noobs<:trolley:1159468147133395025>

ye

I think you're right

You don't need to argue for me😊

Btw Which link do you use to make the model?

what do you mean

I MEAN VOICE MODEL

AI

YOU KNOW

how do i make them?

I train my models locally

if thats what youre asking

That's definitely what I asking

Could you send the link? If you want

https://github.com/Mangio621/Mangio-RVC-Fork

guides for it can be found in <#1159513888199540817>

Thank you so much😭

Dude I did everything I clicked the go web And it says launch share true bla bla I clicked on the local link it gave me, but it does not open😂

It looks like the local link is banned in my country...lol

Tobey Maguire | ENG | 500epochs | RVC_v2 | RVMPE 40k | spider man