Beru (Solo Leveling Arise Dub) --Dataset will be Provided--

[languid kettle]

Anyone willing to do a model for beru from the solo leveling arise game (english dub)? I plan to use with real time voice changer and possibly other things.

I have a dataset folder to send over for anyone willing take this on (Courtesy of Dryesthalo98066).
Note: Its not a lot and I'm aware it might lead to a less than desirable model, but I'm still willing to give it a shot.

Two versions of the dataset can be provided:

Raw Clips: the clips in the data set are split into multiple files for a single line for most of them (the raw ripped files).

**Stitched:**My stitched together edits of the clips to make full sentences as they were meant to be.

[high arrow]

Question, do you know if Beru was added in a recent update to the game? I ripped it a while back in December but I dont seem to have any assets pertaining to Beru

@languid kettle

[languid kettle]

beru was added roughly around dec 19.

[high arrow]

Got it so after I ripped it, Ill rerip it and see if I can get anything new

[languid kettle]

cool, thx

[high arrow]

Yeah

[high arrow]

@languid kettle update so I ripped all the assets and I found some dialogue. It’s not a lot though. I can send it over but it likely isn’t enough for a great model.

[languid kettle]

@high arrow do the assets include what I highlighted in the cutscene vid. I don’t know much about asset ripping but I’m assuming it doesn’t isolate lines from rendered cutscenes. This is assuming that the cutscene is a video and not just live rigged models.

[high arrow]

It has the isolated voice lines from the cutscene

I would send them here if I could

I can just DM them to you

[languid kettle]

Oh, I was under the impression that you would use them. I don’t know how to make the model. My bad if I wasn’t clear or gave the wrong impression

[high arrow]

All good,

I would but I dont have the time arm

atm

You have a dataset though which is the most difficult part in most cases

[languid kettle]

Ok you can still send the data over and if you know anyone who can take over that would be great. If not I’ll just leave the post up and see if if I get any hits

[high arrow]

I dmed the dataset to you

[languid kettle]

Awsome. Thx again

[high arrow]

👍

[languid kettle]

Beru (Solo Leveling Arise Dub) for Real-Time Voice Change (Dataset will be Provided)

[languid kettle]

Beru (Solo Leveling Arise Dub) --Dataset will be Provided--

[high arrow]

Has anyone made the model for you yet?

@languid kettle

[languid kettle]

not yet

[high arrow]

Ok, I’ll train it later today since I got some more time now. Just a heads up though, don’t expect the model to be that great. It might sound relatively ok but with a dataset as small as this it might be a bit wonky.

[languid kettle]

yeah no worries. if it doesnt work i can always find a character with similar voice and a larger data set size. do you want me to send over my stitched files?

[high arrow]

That’s fine, I got my own method of processing. I might have some ideas for some similar sounding characters though. They have the same “vocal effect” but we will see how this model sounds first.

[languid kettle]

ok. also one of the alternate characters i had in mind was savitar from the cw flash. if the pitch of his voice is lowered slightly it might be a good fit

[high arrow]

Yeah, I could see that. The easiest characters will be video game characters since the isolated dialogue can be ripped. I’ll show you the character that I was talking about later.

[languid kettle]

sounds good

[high arrow]

@languid kettle I ended up just using some other characters and mixing and matching dialogue. The result sounds better than I expected. Here https://huggingface.co/Dryesthalo98066/BeruV1/resolve/main/Beru.zip?download=true

It doesnt sound exactly like Beru but it sounds close enough to what I think you were trying to get

[languid kettle]

cool ill test it out. what f0 det. should i set the vc to use.

[high arrow]

rmvpe like standard

[languid kettle]

its very close. if it were little less muffled/ cluttered it would be spot on.

@high arrow dm'img a test clip withe the vc

[high arrow]

Cool, maybe in the future of Beru gets more data it will be better. For now I think that is about the best I can do with the data that I currently have

[languid kettle]

it definitely came out better than i was imagining. who did you mix it with?

[high arrow]

https://www.youtube.com/watch?v=-0IoFX6HOHY

Not all of them, but I took a few characters from this list since they sound very similar

[languid kettle]

Nazguls is a definite match. i feel like had you pitched his lines up slightly it would have worked better. i think the formula that would work best when mixing with beru would be 70% Nazguls and 30% suladan (maybe slightly less). the other voices are very overpowering in their use of effects over the man voice

[high arrow]

Yeah, thats similar to what I did.

[languid kettle]

great work nonetheless. if you're willing to try one more with raising the pitch of the mixed lines that would be great. but overall im satisfied with the current result if you are busy

[high arrow]

I will see what I can do, might be a minute though.