Alice Merton - No Roots, by Neuro. | Neuro-sama Headquarters | Page 1

warm ravine Apr 28, 2024, 5:14 PM

#

#1191795539830919318 message
"Fine, I'll do it myself"

loud vortex Apr 28, 2024, 5:15 PM

#

It was converted against the original voice, correct?

warm ravine Apr 28, 2024, 5:16 PM

#

loud vortex It was converted against the original voice, correct?

yeah

loud vortex Apr 28, 2024, 5:16 PM

#

Lemme try something catSUS

warm ravine Apr 28, 2024, 5:16 PM

#

Want me to drop the original files?

loud vortex Apr 28, 2024, 5:16 PM

#

Nah, I am good

warm ravine Apr 28, 2024, 5:16 PM

#

(I've already sliced and separated the tracks)

loud vortex Apr 28, 2024, 5:16 PM

#

Thanks

warm ravine Apr 28, 2024, 5:16 PM

#

aight

#

another song is on the way

#

one I've requested, too xd

tawdry sinew Apr 28, 2024, 5:38 PM

#

Oh, you posted this here, cool

tawdry sinew Apr 28, 2024, 5:42 PM

#

warm ravine yeah

Is this confirmed information or just what you assume, since you didn't make the model?

warm ravine Apr 28, 2024, 5:42 PM

#

tawdry sinew Oh, you posted this here, cool

That's the guy I got the RVC from btw

tawdry sinew Apr 28, 2024, 5:42 PM

#

And I got it from wispers, I made the dataset

warm ravine Apr 28, 2024, 5:42 PM

#

tawdry sinew Is this confirmed information or just what you assume, since you didn't make the...

It seems I have misunderstood CJMAXiK's question

#

I understood "original voice" as the voice of Alice Merton

tawdry sinew Apr 28, 2024, 5:43 PM

#

I don't quite understand it either, is it referring to Evil's voice or the original singer?

warm ravine Apr 28, 2024, 5:44 PM

#

That's what I said

#

I still have no clue

#

@loud vortex pls clarify

tawdry sinew Apr 28, 2024, 5:44 PM

#

Well, that information would be kinda improtant to know

loud vortex Apr 28, 2024, 5:45 PM

#

Original voice is Merton's

tawdry sinew Apr 28, 2024, 5:45 PM

#

I think that's what it's made from, I have another song that sounds better

#

That was also made by converting from the original voice

#

I can send that one if you want to hear what the model can do

#

Since the configuration Kai is using might be wrong somehow

warm ravine Apr 28, 2024, 5:51 PM

#

Ye, I just found a setting that might change everything

#

once I'm done with the second song, I'll get back to this one

loud vortex Apr 28, 2024, 5:53 PM

#

#

Found some random SVP on the Internet

tawdry sinew Apr 28, 2024, 5:56 PM

#

Is that the same model or a different one?

#

It sounds noisier than this one

warm ravine Apr 28, 2024, 5:57 PM

#

loud vortex Found some random SVP on the Internet

SVP?

loud vortex Apr 28, 2024, 5:57 PM

#

I dunno, but it's from wispers as well. I deleted mine around 10 minutes ago before I found the issue with settings, gotta train it once again

#

Tuning for SynthV

warm ravine Apr 28, 2024, 5:57 PM

#

Also, here's another difference from that option

#

(It took only 15 seconds before)

#

(It's been 11 minutes)

#

maybe I overkilled it

warm ravine Apr 28, 2024, 5:58 PM

#

loud vortex Found some random SVP on the Internet

Holy shit

loud vortex Apr 28, 2024, 5:59 PM

#

Tuning is not that good, but I don't feel like fixing it

#

catSUS

#

And it is noisy because there is a heavy post-production chain involved

tawdry sinew Apr 28, 2024, 6:00 PM

#

The model from my dataset doesn't seem to have noise, at least when used correctly

warm ravine Apr 28, 2024, 6:00 PM

#

I'll try to top that, brb

loud vortex Apr 28, 2024, 6:01 PM

#

Arms race lul

tawdry sinew Apr 28, 2024, 6:02 PM

#

This is some of the model's high-end output

warm ravine Apr 28, 2024, 6:07 PM

#

idk why the pitch is so broken for me lmao

tawdry sinew Apr 28, 2024, 6:10 PM

#

Must be some setting you're getting wrong, the model wors really well with whatever wispers is using

warm ravine Apr 28, 2024, 6:11 PM

#

goddamnit

warm ravine Apr 28, 2024, 6:12 PM

#

loud vortex Found some random SVP on the Internet

What's an SVP though?

loud vortex Apr 28, 2024, 6:13 PM

#

loud vortex Tuning for SynthV

^

warm ravine Apr 28, 2024, 6:13 PM

#

huh

loud vortex Apr 28, 2024, 6:13 PM

#

I didn't use the original Merton's voice track, I generated a different one using SynthV

warm ravine Apr 28, 2024, 6:14 PM

#

Ooh

#

makes sense, then

loud vortex Apr 28, 2024, 6:14 PM

#

Alright, back to training catSUS

warm ravine Apr 28, 2024, 6:15 PM

#

loud vortex Alright, back to training <:catSUS:1157367828035420160>

damn

#

This is bullshit, I just trained Gagarin and brother's already on Mars

#

angery

#

damn

warm ravine Apr 28, 2024, 6:34 PM

#

loud vortex I didn't use the original Merton's voice track, I generated a different one usin...

Welp, I can't get "ABBA - Lay All Your Love On Me" working no matter the settings, so, womp womp ig...

#

Could you make it sometime, should you be free, please?

#

really wanted to hear that song from Evil lol

tawdry sinew Apr 28, 2024, 6:38 PM

#

You can also ask wispers, they seem to know how to use the model correctly. You just need to split out the vocals and lyrics

loud vortex Apr 28, 2024, 6:39 PM

#

I know how to do that catSUS

warm ravine Apr 28, 2024, 6:39 PM

#

catSUS

#

I already disturbed wispers today

#

Once a day is the limit catSUS

tawdry sinew Apr 28, 2024, 6:40 PM

#

Ok then, whatever works

#

The model from wispers using my dataset is probably the current best one anyway

#

Just the high ranges are a bit inaccurate

loud vortex Apr 28, 2024, 6:42 PM

#

My model might be better, but I have to re-train it

#

(I am dumb)

tawdry sinew Apr 28, 2024, 6:43 PM

#

I can give you my dataset if you want, it helps a lot with reducing noise based on what wispers got

loud vortex Apr 28, 2024, 6:43 PM

#

warm ravine Could you make it sometime, should you be free, please?

My GPU is busy rn, so take these instead

tawdry sinew Apr 28, 2024, 6:43 PM

#

tawdry sinew I can give you my dataset if you want, it helps a lot with reducing noise based ...

I spent like a moth making it

#

And half a spacebar

warm ravine Apr 28, 2024, 6:43 PM

#

loud vortex My GPU is busy rn, so take these instead

aaaah

#

thanks :D

warm ravine Apr 28, 2024, 6:46 PM

#

loud vortex My GPU is busy rn, so take these instead

what GPU do you use btw?

warm ravine Apr 28, 2024, 6:46 PM

#

tawdry sinew And half a spacebar

lol

loud vortex Apr 28, 2024, 6:46 PM

#

3070 Ti

warm ravine Apr 28, 2024, 6:46 PM

#

nice

tawdry sinew Apr 28, 2024, 6:46 PM

#

I have a 4070Ti

warm ravine Apr 28, 2024, 6:46 PM

#

2070 Super here

#

3 generations in one thread

#

all 70s haha

tawdry sinew Apr 28, 2024, 6:47 PM

#

And all xx70 class

warm ravine Apr 28, 2024, 6:47 PM

#

^

tawdry sinew Apr 28, 2024, 6:47 PM

#

And two are 70Tis

#

Well, even if the model wispers made is not the most accurate for now, it's certainly the least noisy, at least as far as I can hear. I think wispers also said they're still gonna be adding more Evil singing data to the training dataset for the model to increase accuracy and range, with my dataset to stabilize it and reduce noise

#

Since my dataset is quite a large speech dataset, so apparently it's good for that

loud vortex Apr 28, 2024, 7:01 PM

#

Well, wish me luck

warm ravine Apr 28, 2024, 7:01 PM

#

good luck o7

tawdry sinew Apr 28, 2024, 7:01 PM

#

loud vortex Well, wish me luck

Are you still using the same dataset as before?

loud vortex Apr 28, 2024, 7:02 PM

#

Yes, I just need to bring my model back

tawdry sinew Apr 28, 2024, 7:02 PM

#

My dataset apparently helps as an additional thing to stabilize things

#

So I can give you it if you want it

loud vortex Apr 28, 2024, 7:03 PM

#

Not now

tawdry sinew Apr 28, 2024, 7:03 PM

#

Ok then, let me know if you want it at some point

#

I'll continue working with wispers to make our model better

#

Currently it apparently only has like 20 songs on top of my 1180 or something files of speech

loud vortex Apr 28, 2024, 7:05 PM

#

My set is 35 voice tracks from songs

tawdry sinew Apr 28, 2024, 7:07 PM

#

Well, wispers says they'll be adding more to ours to improve the model, especially for higher ranges

loud vortex Apr 28, 2024, 7:07 PM

#

I know (he's in my DMs catSUS )

tawdry sinew Apr 28, 2024, 7:08 PM

#

Makes sense

#

Well, we'll see who manages to make the best model, but ours seems to be really good when used correctly, so maybe adding in more song data will improve it above what yours can do while having my large speech dataset to stabilize it and reduce noise, or at least that's what wispers said it does

#

The noise from what I've heard from your model is not really good

#

Especially for what I plan to use the model for

#

Which is DIVA modding

warm ravine Apr 28, 2024, 7:13 PM

#

only issue I have rn is "sha-ring" and "waa-sting", "emo-hion"

tawdry sinew Apr 28, 2024, 7:14 PM

#

From your output or something else?

warm ravine Apr 28, 2024, 7:14 PM

#

ye

warm ravine Apr 28, 2024, 7:15 PM

#

loud vortex My GPU is busy rn, so take these instead

and those

tawdry sinew Apr 28, 2024, 7:16 PM

#

Well, who knows how much the model will improve with more songs fed in

loud vortex Apr 28, 2024, 11:28 PM

#

This is promising

warm ravine Apr 29, 2024, 1:32 AM

#

loud vortex This is promising

I tinkered with that a bit and made this:

warm ravine Apr 29, 2024, 1:50 AM

#

Ain't that nice
https://docs.google.com/spreadsheets/d/1tvZSggOsZGAPjbMrWOAAaoJJFpJuQlwUEQCf5x1ssO8/edit?pli=1#gid=0

Google Docs

VTuber AI Model Tracking v2

RVC

Virtual YouTuber AI Model Tracking
Spreadsheet made and maintained by Kit Lemonfoot and Bartezes
Submission Form
VTuber Name,Organization,Epoch Count,Dataset Length,RVC Version,D...

#

There's over 600 entries

#

given, there's like 2-5 of each voice

#

anyway, eepy time

tawdry sinew Apr 29, 2024, 3:29 AM

#

Should we get our model listed there at some point?

loud vortex Apr 29, 2024, 10:30 AM

#

I prefer to have my models exclusive to me.

tawdry sinew Apr 29, 2024, 10:34 AM

#

I prefer to share everything possible, which is why I put the model on my public FTP server

#

The one wispers made from my dataset

#

So even if your is better, ours is more accessible

#

And I think ours could become as good or better than your model with enough data

loud vortex Apr 29, 2024, 10:41 AM

#

I like this competitive nature catSUS

#Alice Merton - No Roots, by Neuro.