well 1. yeah i know what you mean thats what i always do yet still causes immense lag for heavy games
ill take a look at lightning.ai
#✨│ai-help
supple trail
overall i just did this cause i enjoy coding : p
im not gonna pr it cause im still in the process of testing this solution i may not have came across bugs yet but im still testing
low shard
help me
this is a General AI Discord Server, Please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
lone abyss
RVC Apollio colab stopped working, what do I do?
the public link doesn't load as it gives an error about that..
lone abyss
ngrok gave me the same issue
I didn't touch anything, I have no idea what is "launch" I tried to find it
nocturne mural
ngrok gave me the same issue
lone abyss
Just click the link and enter the password you've been given
oh thank you
it's kinda slow though, should I wait for the other options to be back up if I wanna train a new model?
nocturne mural
it's kinda slow though, should I wait for the other options to be back up if I w...
viral mason
who?
check here https://discord.com/channels/1159260121998827560/1175430844685484042 and here https://voice-models.com I have zero idea who that person is
if u can't find a model of her you could make one yourself or commission a free one
for making a model I'd suggest applio, there's a guide for it yea
-rvc
viral mason
should have all the info you need ^^
you're welcome! if you have questions on anything ping the @ helpers role
hardy yew
who?
if i'm not mixing stuff, i think that's the popular minecraft youtuber girl who turned out to be a guy with voice changer
so basically catfishing final boss
hardy yew
there was a discussion about this on vonovox discord a couple days back
otherwise the name wouldn't ring a bell xD
viral mason
I have horrible memory sometimes so some stuff just goes in and gets turned to ash
hardy yew
well, this is the kind of knowledge i wouldn't mind forgetting xD
pallid crater
Can you give me the Weights server?
viral mason
the site is shut down there isn't much of a need for it
plus can't send invites to discord servers here anyways
should be on the site if it still even works
surreal salmon
Guys, i am probably blind, but can you tell me if they are something in vovonox where you can here yourself like, i don’t even know 😭
hardy yew
No, not directly in vonovox. But you can go to windows sound settings, properties of your VAC and enable "Listen to this device" so that you can hear what passes through it
surreal salmon
No, not directly in vonovox. But you can go to windows sound settings, propertie...
Oh ok tysm, but are they something better than vovonox?
For voice changer *
rtx 3060
hardy yew
Nah, vonovox is the best for now
surreal salmon
Ok ty
drowsy hinge
can i make the rvcaicovermaker so that it covers the whole song and not only the vocals?
buoyant pond
Hello everyone, I'm looking for the RVC for the “Alex” voice—that's what it's called on Eleven Labs. Does anyone know where I can find it?
crystal crystal
What do I use to run the voice models?
oak wadi
can someone help create and train an rvc model for lakyus alvein dale aindrs of overlord
shell marlin
Did they just nuked the Weights Discord Server even if they stated that they won't do it?
finite light
Guys, is Replay safe for PC?
viral mason
<@&1159293140440723499>
pastel sand
Does anyone know if it's possible to try Seedance 2.0 for free?
velvet meteor
does this work on intel the voice changer
finite light
I definitely need better and strong laptop to have more space... to install Replay.
mellow wigeon
Hi, I need some assistance on my build. apparently I don't have pipeline working
bold sphinx
can anyone tell me how to use any model and where to use it
candid iron
How do I get Beatrice models?
low shard
does this work on intel the voice changer
Elaborate:
- your PC GPU
- your PC os
- what you're trying to do, like TRS, e Girl Trolling / Catfishing, AI Covers, Roleplaying
- the tutorial link used
Don't ping mods,
this is a General AI Discord Server, Please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
low shard
I definitely need better and strong laptop to have more space... to install Repl...
What's your PC GPU and OS?
low shard
Hi, I need some assistance on my build. apparently I don't have pipeline working
this is a General AI Discord Server, Please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
low shard
can anyone tell me how to use any model and where to use it
this is a General AI Discord Server, Please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
low shard
How do I get Beatrice models?
No one trains beatrice models because they are lower quality than RVC, Please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
low shard
Did they just nuked the Weights Discord Server even if they stated that they won...
It's not nuked
low shard
can someone help create and train an rvc model for lakyus alvein dale aindrs of ...
What's your PC GPU and OS?
low shard
What do I use to run the voice models?
this is a General AI Discord Server, Please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
bold sphinx
this is a General AI Discord Server, Please elaborate:
- your pc gpu
- your pc o...
i want to know how to use these models and i just want to test it
keen mica
in uvr whats the best background song remover that doesn't remove sound effects and vocals
low shard
i want to know how to use these models and i just want to test it
there isn't a single program for everything, you need to elaborate
proud osprey
https://www.tiktok.com/@picklerickursion/video/7602691866852707606 - can somebody help me to find what voices are they using for the video?
bold sphinx
there isn't a single program for everything, you need to elaborate
is there any guide for it
low shard
is there any guide for it
there are multiple guides, you need to elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
sharp jungle
there are multiple guides, you **need** to elaborate:
- your pc gpu
- your pc os...
Nick can u give me the colab/kaggle notebook which directly outputs an song with RVC used and I don't need to manually extract lead vocals etc
There used to be one...if i remember correctly
low shard
Nick can u give me the colab/kaggle notebook which directly outputs an song with...
it's best to firstly check if you can do it locally before using cloud options, what's your pc gpu and os?
sharp jungle
it's best to firstly check if you can do it locally before using cloud options, ...
I have medium specs to do it locally fr sure... but i just prefer cloud since i do many songs in batches... it's not regular thing
I do run regular RVC locally fr inferencing currently
low shard
I do run regular RVC locally fr inferencing currently
could you please tell your pc gpu and os? the same thing that exists on cloud exists also locally
winged adder
.
Hi everyone,
I’m on a quest to find the absolute speed king in the Mixture of Experts (MoE) category, specifically models under the 9B–14B parameter range.
The Use Case
I’m building a local sentence rewriter. The goal is to take a "bag of words" and transform them into syntactically and semantically perfect sentences.
- Example: Input: cat, fish, lake ➔ Output: The cat eats the fish by the lake.
The Technical Setup - Hardware: Dual GPU (16GB VRAM total) running via Vulkan.
- Platform: lm.
- Constraint: I am not looking to do a custom fine-tune with Wikipedia data; I need something high-performing out of the box.
The "HellaSwag" Problem
In my experience, small models usually fall apart on HellaSwag (common sense reasoning). Most sub-14B models lose significant syntactical "intelligence" because they lack the training depth. I need a model that maintains elite HellaSwag scores while remaining an MoE for that snappy inference speed.
L
You can see the leaderboard I'm tracking here:
:https://huggingface.co/spaces/sparse-generative-ai/open-moe-llm-leaderboard
What is the fastest English MoE under 14B that doesn't sacrifice HellaSwag reasoning for speed? If it’s on that leaderboard and runs like lightning on Ollama/Vulkan, I want to hear about it.
Looking forward to your suggestions
pastel sand
Any idea where I can test Seedance 2.0?
sharp jungle
could you please tell your pc gpu and os? the same thing that exists on cloud ex...
3050 and windows 11
I wanted the one which when given an input song... automatically pr-prcrss isolates lead,removes reverb etc applies rvc and then joins it with the inst with bgv with applied settings of loudness reverb etc
The colab was Hina something if i recall it properly
hallow thistle
3050 and windows 11
"Applio RVC" can work with that GPU locally, but Applio RVC is manual. "Music Source Separation Training (MSST)" is a type of UVR5 that is developed to separate audio files in batch, also exists as local program but there's Colab notebook as well.
sharp jungle
"Applio RVC" can work with that GPU locally, but Applio RVC is manual. "Music So...
Ya i liked the colab one cause with just one cell run and install i can do multiple songs in one go...also doing locally takes space and my gpu is a bit slower than t4
Also i hate using uvr5/mvsep
It's like too much mannual hardwork
Btw can u link me that colab if it still works and if y'all have the link
Without dependency issues ofc
hallow thistle
Big excuses for non-techy users but that's about it. 
This is MSST Colab notebook that I often use to separate audio stems in batch, still manual to select models and settings, not really a big code cell full automated one-click thing. https://colab.research.google.com/drive/1IC6Q1hLF55_tK6mhky0SWYKGVF9T5WsY
viral mason
This is ai help
bitter forum
why is not ai helping me then
viral mason
Bc a robot is dumb :p
Anyways for Nvidia you should use Vonovox, what do you plan on using it for btw just curious
bitter forum
Anyways for Nvidia you should use Vonovox, what do you plan on using it for btw ...
sounding like valorant characters
i wanna sound like omen
no gif perms
viral mason
He looks pretty cool
Here's the two downloads you'll need, for Vac lite (the second one) extract, then run setup64
And for Vonovox extract then run start, it's a .bat file
@bitter forum
bitter forum
Here's the two downloads you'll need, for Vac lite (the second one) extract, the...
i have venovox downloaded what is vac lite tho
fiery raptor
does this mean i gotta download the models here (i downloaded the applio local run for context)
viral mason
i have venovox downloaded what is vac lite tho
It's a virtual audio cable, it basically will connect Vonovox to games or discord so people can hear the voice changer
bitter forum
It's a virtual audio cable, it basically will connect Vonovox to games or discor...
ohh i alr have venovox setted up then but how do i fix the delay
it like has delay and stutters
viral mason
Just lowering the block size, .30 should be fine but some people use .25
bitter forum
you mean chunk?
viral mason
Same thing but with a different name on Vonovox, all you can do is lower or raise it tho as the other stuff was set to the best settings internally
fiery raptor
do i gotta press the run-applio.bat everytime i wanna use it
viral mason
does this mean i gotta download the models here (i downloaded the applio local r...
This section is for testing models using audio samples
viral mason
do i gotta press the run-applio.bat everytime i wanna use it
Most likely, I'm unsure how it works locally
To download models to applio tho go to the download section inside it and upload a model in a zip file (both index and pth file in the 1 zip)
Then just refresh on the inference section and it should be there
viral mason
you mean chunk?
Btw make sure to set the pitch of all models that aren't female ones to 0 unless they have a high pitched voice because even tho I have told the creator multiple times to set the default to 0 he hasn't
fiery raptor
Most likely, I'm unsure how it works locally
okay i think i did manage to get the model working but whats going onm here
it just not loading
viral mason
You still have the command window open right?
fiery raptor
yea
viral mason
Alr
Tbh I'm not experienced enough in the local version of applio maybe a helper like Namari would know
I'm sorry
fiery raptor
its alr
fiery raptor
if i was experienced enough i wouldnt ask for help
fiery raptor
ohh
bitter forum
viral mason
Oh
fiery raptor
wait i can only find the og file
bitter forum
what does that even mean
nocturne mural
wait i can only find the og file
Then some error has occurred. Could you show me what the console is displaying?
bitter forum
Could you screenshot your audio settings rq
oh that matters
idk i just download random sample and set those randomly
😭 \
viral mason
Huh?
bitter forum
ima look for omens one
viral mason
No I mean like what is your input and output
viral mason
My bad lol
bitter forum
viral mason
Uh
viral mason
Did you have a separate voice changer running before opening Vonovox?
That could have affected it
Close everything rq then open just Vono
fiery raptor
@nocturne mural dont go buddy i need ur help 
nocturne mural
It looks like the .index file for the model you're trying to use is corrupt or invalid for inference. Maybe you could just go to the logs folder, try deleting it, hit the refresh button, and try the inference again.
nocturne mural
<@989772388508000306> dont go buddy i need ur help <:cat_seriously:1366692376575...
I'm using a translator.
fiery raptor
I'm using a translator.
you dont speak english?
fiery raptor
well you are helpful
nocturne mural
nocturne mural
patent trellisBOT
zinc frost
<@&1159293204038955078> <@&1159293140440723499> can you advise me what version of wokada I have to use with windows 10?
tame oracle
<@&1159293204038955078> <@&1159293140440723499> can you advise me what version o...
that depends on your gpu and what are you trying to do with it
wanton gulch
<@&1159293204038955078> <@&1159293140440723499> can you advise me what version o...
Don’t ping mods for that
low shard
3050 and windows 11
low shard
ima look for omens one
are you looking for women voice model to e girl troll / catfish?
low shard
<@&1159293204038955078> <@&1159293140440723499> can you advise me what version o...
don't ping mods for ai help
use https://discord.com/channels/1159260121998827560/1489232949176959088
latent knoll
i trained my model with mangio-rvc but i didnt get a pth file what do i do
sharp jungle
This is MSST Colab notebook that I often use to separate audio stems in batch, s...
Ty
sharp jungle
check https://docs.aihub.gg/rvc/local/aicovermaker/
Ty
viral mason
i trained my model with mangio-rvc but i didnt get a pth file what do i do
Mangio is really old, how come you did that?
sharp jungle
i trained my model with mangio-rvc but i didnt get a pth file what do i do
I think it somehow failed
latent knoll
Mangio is really old, how come you did that?
did what
viral mason
Trained with mangio over rmvpe
latent knoll
viral mason
?
latent knoll
thats the logs stuff that opened on my pc
viral mason
Oh
latent knoll
if it failed or something
sharp jungle
if it failed or something
It did definitely
viral mason
Are you training your model with applio?
latent knoll
no i downloaded mangio
sharp jungle
Are you training your model with applio?
That is not applio
latent knoll
viral mason
no i downloaded mangio
You should be using Applio, it's the standard to make models now, that looks super old
Like 2023 old
latent knoll
download link?
viral mason
-rvc
latent knoll
and tutorial on how to make models
viral mason
Any YouTube tutorials are outdated you'll need to read the docs
sharp jungle
download link?
U can also just use the regular RVC too, i like that...but applio has way more features
latent knoll
applio docs?
sharp jungle
applio docs?
Or just use kaggle
viral mason
Yea the second link is what you should use
latent knoll
whats kaggle
viral mason
Or just use kaggle
Kaggle is a site you can use applio on
latent knoll
i got 0 coding or whatever knowledge
sharp jungle
Kaggle is a site you can use applio on
U can train model there too
latent knoll
oh i had kaggle
sharp jungle
whats kaggle
It's an online web based platform to run codes, create notbooks etc
latent knoll
that didnt let me train models at all
sharp jungle
Kaggle is the most eaasy one
viral mason
True I currently use it
sharp jungle
In local there u have to download, if ur gpu is mid.. training is slow asf
latent knoll
so i should use kaggle for my best one?
latent knoll
In local there u have to download, if ur gpu is mid.. training is slow asf
i got AMD RX 7600
sharp jungle
so i should use kaggle for my best one?
To train yes ..but fr inferencing/running the model...just use local
sharp jungle
i got AMD RX 7600
Vram?
latent knoll
so i should use kaggle for my best one?
first*
sharp jungle
Settings
I mean just use kaggle to train... it's fast
U get more than enough time to train 1000 epochs
viral mason
I mean just use kaggle to train... it's fast
He would need to know how to use Kaggle
sharp jungle
is it here
8 gb vram
latent knoll
btw my data is 1:30 minutes will it be good
sharp jungle
btw my data is 1:30 minutes will it be good
I mean it's too low...so don't expect good model
But it'll train super fast
latent knoll
I mean it's too low...so don't expect good model
how much sould i use
another question i wanna make a rapper voice model but he uses a lot of adlibs and i cant get rid of them, will that mess up my thingy
viral mason
btw my data is 1:30 minutes will it be good
Eh.. I have made models with really low data but they usually just turn out meh
I know what I'm doing though
Been training for years
sharp jungle
how much sould i use
Min 15-20
But the more the better i like my 40-1hr models
sharp jungle
another question i wanna make a rapper voice model but he uses a lot of adlibs a...
Use mvsep karaoke models
To get lead vocals
latent knoll
the thing is the rapper uses a different voice for most of hjis songs
yea i did but the adlibs are very loud so i guess it thinks its main vocals
sharp jungle
the thing is the rapper uses a different voice for most of hjis songs
Oh like different eras?
latent knoll
Oh like different eras?
no just sometimes he screams at the mic and sometimes he talks nonchalantly
sharp jungle
Tbh it's preferred to use consistent voice/ same type of voice...idk how it'll turn out in ur case
viral mason
Min 15-20
But the more the better i like my 40-1hr models
It's not totally true, it's more quality or quantity but having around 10-15 mins + is good yea
latent knoll
Tbh it's preferred to use consistent voice/ same type of voice...idk how it'll t...
i am
thats why the data is so short
sharp jungle
It's not totally true, it's more quality or quantity but having around 10-15 min...
Ya i meant that...like good quality dataset is very important
Like studio stems
viral mason
True yea
sharp jungle
Without reverb
latent knoll
lee drilly
viral mason
I always remove reverb, then echo
latent knoll
yo bro this is NOT simple
sharp jungle
Just see how it turns out...i mean in ur case the model should train fairly fast
-kaggle
viral mason
yo bro this is NOT simple
I'll jump on my pc and record how to do this hold on
It's super simple
latent knoll
I'll jump on my pc and record how to do this hold on
thank you man
sharp jungle
I'll jump on my pc and record how to do this hold on
Nice
latent knoll
1st link?
latent knoll
oh alright
sharp jungle
Sign up in ngrok
latent knoll
i am
sharp jungle
Learn all this one time...it may look hard but then u can create as many models u want
latent knoll
okk
viral mason
Also u need ngrok token
Not really
sharp jungle
Not really
I know it's optional but i like it
latent knoll
will it make anything easier / my model better?
sharp jungle
will it make anything easier / my model better?
Not really but in case the session times out ...ur files will still be saved in ngrok
Temporarily tho
Also u can see the tensor board log graphs so the model doesn't overtrain
sharp jungle
isnt the more i train the better
No
At one point It'll reach a point where the more u the train the model, the worse the model will become
latent knoll
yo i know i got a lot of questions but like i also got 4 studio sessions from a rapper and i wanna train a model using them, but theres him whispering to himself and some background noise there, will it stil turn out good
latent knoll
At one point It'll reach a point where the more u the train the model, the worse...
how do i know where that point is?
viral mason
isnt the more i train the better
the best solution to seeing how good or bad your model is at any time is to just use samples like raw vocals of a song or someone talking'
sharp jungle
how do i know where that point is?
Applio has a overtraining feature to prevent it
viral mason
test each timee
viral mason
Applio has a overtraining feature to prevent it
overtraining detector basically doesn't work
sharp jungle
overtraining detector basically doesn't work
I mean i never used it..so idk...i always check my tensoboard log graphs... specially the loss graphs
viral mason
yo i know i got a lot of questions but like i also got 4 studio sessions from a ...
if it's just microphone noisee it's fine
latent knoll
hes like whispering the lyrics hes going to say and the beat is leaking a bit
hes the door closes/opens sometimes
viral mason
I mean i never used it..so idk...i always check my tensoboard log graphs... spec...
tensorboard really doesn't help tbh, I stopped using it year or so ago and have been getting better models by just listening to when my model starts sounding good or loses quality/sounds flat ect
I test every 5 or 10 epochs
sharp jungle
I test every 5 or 10 epochs
yaa that's a good way to know too..cause some times it's personal choice which epoch sounds best to someone
viral mason
true yea
latent knoll
@viral mason did you record the thing you said?
sharp jungle
Why is there no mods here to help u lol
viral mason
<@1023278814752677918> did you record the thing you said?
currently still getting all the footage I need
viral mason
Why is there no mods here to help u lol
most likely the helpers are asleep or are at school or work
I don't have a life currently so I'm always here
rvc is the one that sticked
don't know what svc is, what is that?
what is better than rvc? I feel like I wouldve heard of it
currently I really like Good Kid
use Bubbly, it's pretty good
spring kraken
someone please dm me that knows about good and intercative nice desinged pdf or something like that and knows about automation please
viral mason
<@1023278814752677918> did you record the thing you said?
almost done
latent knoll
almost done
alrightt
viral mason
should send it to my dms instead in case of copyright rules here
burnt tangle
almost done
One message removed from a suspended account.
viral mason
yea
latent knoll
what did this do?
low shard
One message removed from a suspended account.
what's your pc os?
are you trying to do TTS, AI Covers or E Girl Trolling / Catfishing?
viral mason
holy shit this sounds honestly really like him
burnt tangle
what's your pc os?
are you trying to do TTS, AI Covers or E Girl Trolling / Cat...
One message removed from a suspended account.
One message removed from a suspended account.
viral mason
One message removed from a suspended account.
Joe Swanson you mean?
low shard
One message removed from a suspended account.
yeah either vonovox or wokada tg develop would work fine for you
burnt tangle
One message removed from a suspended account.
low shard
i think that's joe btw
burnt tangle
yeah either vonovox or wokada tg develop would work fine for you
One message removed from a suspended account.
One message removed from a suspended account.
spring kraken
someone please send me that knows about good and intercative nice desinged pdf or something like that and knows about automation please
viral mason
is the software free?
low shard
One message removed from a suspended account.
-realtime
low shard
someone please send me that knows about good and intercative nice desinged pdf o...
wdym
viral mason
that person has asked that twice now
darn, it actually seems like a nice program
does it have a website?
viral mason
if you're going to use vonovox use the download I sent to the other user as the beta is much better than the most recent non beta release
spring kraken
cant i ask this guys?
burnt tangle
if you're going to use vonovox use the download I sent to the other user as the ...
One message removed from a suspended account.
spring kraken
cant i ask this guys?
its an ai hub discord?
viral mason
alrightt
this is a somewhat ok guide to use Applio on Kaggle for now, really bland since it's mostly silent
doesn't show how to isolate audio as that will be seperate
the video is already long enough lol
latent knoll
alrighttt thxxx
viral mason
lol
latent knoll
i didnt download ngrok tho
viral mason
here's the basic audio separation guide, just do this for all audio you're using, for seperating bg music and vocals I use the one shown in the video
latent knoll
here's the basic audio separation guide, just do this for all audio you're using...
as a producer i already know it lol thank you tho
viral mason
plus this stuff, bg vocal separation, reverb removal, then echo removal
viral mason
i didnt download ngrok tho
don't need it, you'll just switch ngrok to Gradio + LocalTunnel like in the video
latent knoll
and i need to do this everytime i wanna make a model? im a minute in
viral mason
every time yea
btw important note for kaggle specifically
the batch size is doubled by default so 2 would be 4
and so on
latent knoll
idk nothingabout making models ill have to leanr all of this
viral mason
batch size is dependent on audio length but I mostly would do either 4 or 8
latent knoll
batch size is dependent on audio length but I mostly would do either 4 or 8
is there a table i think its called showing the batch size for every length
viral mason
is there a table i think its called showing the batch size for every length
wdym?
latent knoll
like a
like this i guess
but instead of height/weight its length/batch size
i know
@viral mason how do i know how much epoches to use?
viral mason
is there a table i think its called showing the batch size for every length
not really no, just use either 4 or 8, they're the most stable from my experience
viral mason
<@1023278814752677918> how do i know how much epoches to use?
test them starting at around 20 epoch, then go until it sounds good to you
latent knoll
wont it take very long
viral mason
maybe, but it will be the best way to tell if your model is good or bad
you can jump around from like 20 to 100 if you want
don't have to test each one individually
I save every 5 btw since it's better to save every 5 or 10
latent knoll
wait cant i just save every 5 epoches and test each one
viral mason
yea
latent knoll
@viral mason is this normal?
viral mason
looks normal for kaggle yea
just wait a bit to see if it starts generating the epochs
latent knoll
yea it did
i just do this everytime i wanna make a model?
and can i open applio without doing all of this?
viral mason
every time yea
viral mason
and can i open applio without doing all of this?
wdym?
latent knoll
without going into kaggle and importing the applio notebook and stuff
viral mason
you'd have to do that as well
latent knoll
oh ok thank you
viral mason
you're welcome
latent knoll
@viral mason hey bro im tryna make another one bcus the first one came out bad but no file is unclickable
viral mason
<@1023278814752677918> hey bro im tryna make another one bcus the first one came...
you selected the wrong applio
viral mason
think u missclicked on the website one not the first one
viral mason
is the fl studio built in deverb good enough
I dunno never used it
latent knoll
which one do u use? mvsep?
viral mason
yea
latent knoll
oh ok
latent knoll
@viral mason can i get help
viral mason
<@1023278814752677918> can i get help
did you swap the ngrok for gradio
I think you should watch the whole video again just in case
viral mason
.
latent knoll
did you swap the ngrok for gradio
I realized after a few minutes of starting at the problem
latent knoll
@viral mason yo whats the best thing to use when i wanna make stuff with the model and notjust create it
light wave
does anyone knows, why the voices are choppy and not fluid when using voice changer client demo
round magnet
@viral mason How do I get permission to do things? I've been on calls for over 2 hours and still haven't received permission.
viral mason
<@1023278814752677918> yo whats the best thing to use when i wanna make stuff wi...
wdym
viral mason
<@1023278814752677918> How do I get permission to do things? I've been on calls ...
you have to chat to lvl up
round magnet
I can't share my screen, etc.
Yes, but I've already spent more than two hours on calls with people and still nothing. I've also logged in here on other days and nothing.
alr
light wave
does anyone have tips on how to make the voice not choppy anymore
latent knoll
does anyone have tips on how to make the voice not choppy anymore
wdym by choppy
light wave
wdym by choppy
like its not fluid when using any model its choppy
latent knoll
like its not fluid when using any model its choppy
show example
storm agate
need help with w-okada
I got everything set up according to a guide but when I talk I still sound the same
low shard
need help with w-okada
please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
hallow thistle
<@&1159293204038955078> <@&1159293140440723499> can you advise me what version o...
Next time, only ping "helper" for help about a program. You would ping moderator only if there's a spam bot here.
lyric star
-colab
trim wind
is there any rvc programs for macbook air 2020 (m1)?
low shard
is there any rvc programs for macbook air 2020 (m1)?
do you want to make ai covers locally?
you could try https://docs.aihub.gg/rvc/local/applio/#linux--macos, but i'm not sure if it will work
trim wind
do you want to make ai covers locally?
you could try https://docs.aihub.gg/rvc/...
mainly as an exe program. i don't got patience to set it up. just one that is easy and accessible.
low shard
mainly as an exe program. i don't got patience to set it up. just one that is ea...
you could try Applio, I'm not sure if it uses the Mac GPU Backend MPS ( @simple ore correct me if im wrong), but be aware that it's not as easy as just installing an app from an app store, Open Source AI Programs are usually not as 1 click, especially when Apple makes it harder to download any app not from the App Store
you could also try https://www.weights.com/replay, but i'm not sure if it still works, and it's surely not getting future updates since Weights shutdown
trim wind
you could also try https://www.weights.com/replay, but i'm not sure if it still ...
i hope it does because i liked weights and it was easy to use compared to other rvc programs.
low shard
i hope it does because i liked weights and it was easy to use compared to other ...
i fear that the servers where it downloaded the things are down, you could try and let me know
Honestly, I would suggest Applio more, since it's going to get future updates and using old unmaintained programs might not work forever
trim wind
i fear that the servers where it downloaded the things are down, you could try a...
can't you just get the voice models from elsewhere?
low shard
can't you just get the voice models from elsewhere?
#1175430844685484042 , but I didn't mean voice-models, I mean Internal Components needed for it to work, I remember it did download thins from weights servers but I don't remember much
the only way to know is to try it
else, you could use cloud (remote pc gpu) ways, meaning you're limited to their free cloud gpu or you'd need to pay
(no one would give you their GPU for free forever, else they'd go bankrupt like Weights did)
trim wind
oh okay.
low shard
so your options are either try locally with your hardware, or use the free limited cloud ways
if you want suggestion for cloud ways you can ask me and I can send you
trim wind
not saying it has to be free but not overpriced either.
low shard
not saying it has to be free but not overpriced either.
for cloud ways, there are different cloud platforms to run each program
check https://docs.aihub.gg/rvc/cloud/applio-cloud/
or if you only want the easiest way possible to make ai covers like weights did, check https://docs.aihub.gg/rvc/cloud/aicovermaker-cloud
trim wind
i will try your suggestions. thank you.
low shard
i will try your suggestions. thank you.
you're welcome
fossil reef
Can I summon an AI Model with the drive link?
I can´t upload mine's to Hugging Face.
hallow thistle
What is your PC GPU? Did you follow any tutorial or guide before? And what would you use the voice changer for?
viral mason
Can I summon an AI Model with the drive link?
I can´t upload mine's to Hugging ...
just upload it
there shouldn't be anything stopping you
low shard
please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
west vapor
any seedance 2 users here that can let me know if its possible to get the video toinclude the actual audio ref, trying ot do music video and it just writes a new song instead of using the one i tell it to lipsync
edgy violet
Where can I find the removed voice models after weights.gg got shutdown?
low shard
Where can I find the removed voice models after weights.gg got shutdown?
#1175430844685484042 else rip
edgy violet
<#1175430844685484042> else rip
The RJ one is not there.
low shard
The RJ one is not there.
then either #1159289738314919936 or make one yourself
there's no archive of all models that were on weights, it would be wayy too big, weights.com was the archive itself
grim dock
nothing is working when i turn it on
obtuse summit
yo
can any1 help me? im trying to make my rvc model for the forth time....
nvidia 5060 (8gb vram) with a ryzen 5 5600x, 32gb of ddr4 ram, and i have w11 home with 1tb storage, 2TB counting my external drive
my dataset is clean and 3m long
i can upload my dataset but its in italian, im trying to use rvc as an alternative to voicemod, im not spending 70$ for a voice changer
there's nothing in the docs
viral mason
i can upload my dataset but its in italian, im trying to use rvc as an alternati...
the voice changers are free
use Vonovox
obtuse summit
they are not, community voices are paid
obtuse summit
use Vonovox
as the rvc?
viral mason
are you wanting to train a voice, use an rvc voice changer or both?
obtuse summit
i need to create my rvc model and use it
viral mason
for training use Applio
obtuse summit
im downloading applio, i used that for the past 3 tries
viral mason
i need to create my rvc model and use it
for using any voice from here or made by you use Vonovox
it's free
obtuse summit
there's not any N italian rvc model so :/
obtuse summit
for using any voice from here or made by you use Vonovox
for using the rvc i used applio as W'okada didn't work
i never tried vonovox
for applio should i download the zip or use dione
viral mason
there's not any N italian rvc model so :/
if you get the italian dub of Murder Drones and isolate his voice should make a decent model
obtuse summit
i can't upload mt dataset, but its clean and no noise, used capcut and audacity to edit it
viral mason
for applio should i download the zip or use dione
your pc seems like it will run Applio fine locally tho I have no experience locally
-rvc
viral mason
i can't upload mt dataset, but its clean and no noise, used capcut and audacity ...
nice!
obtuse summit
my pc runs almost everything, even ai models that don't run on my pc, like a 14b that uses 10gb of vram but i only have 8
so what should i do
im installing applio in the meanwhile
its kinda stuck...
viral mason
it's stuck?
obtuse summit
yes its stuck at 91%
its not installing
dione isn't installing applio
Starting script 'Applio' on 'C:\Dione\apps\Applio'
All required dependencies are installed.
INFO: Starting step Starting Applio
INFO: Using virtual environment: env with uv
INFO: Changed working directory to: C:\Dione\apps\Applio\applio
OUT: Traceback (most recent call last):
File "C:\Dione\apps\Applio\applio\app.py", line 6, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'
"call "C:\Dione\apps\Applio\env\Scripts\activate" && uv run python app.py && call "C:\Dione\apps\Applio\env\Scripts\deactivate.bat"": import gradio as gr
ModuleNotFoundError: No module named 'gradio'
OUT: import gradio as gr
ModuleNotFoundError: No module named 'gradio'
ERROR: Failed in default startup: Error: Traceback (most recent call last):
File "C:\Dione\apps\Applio\applio\app.py", line 6, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'
now it has ANOTHER error.
im using manual...
viral mason
I made a short tutorial on how to use it
obtuse summit
im istalling it with the manual install, it should work, idk why dione doesn't want to install it
viral mason
hmm
I'm not familiar with Dione but is there a way u could install it not using that?
I doubt that's causing the errors tho
obtuse summit
yes, manual install with hugging face
wait, i didn't update dione, thats why it crashed, and i didn't download the .exe but i put it from microsoft :/
Downloading script from dioneapp/official-scripts
Script downloaded successfully.
All system requirements are met.
Installing dependencies: git, uv, build_tools
Downloading git for windows (amd64) using URL method...
Running command: tar -xf C:\Users\liamv\AppData\Roaming\dione\bin\temp\git-windows-amd64.zip -C C:\Users\liamv\AppData\Roaming\dione\bin\git
git installed successfully
Downloading uv for windows (amd64) using URL method...
Running installer for uv...
Running command: tar -xf C:\Users\liamv\AppData\Roaming\dione\bin\temp\uv-windows-amd64.zip -C C:\Users\liamv\AppData\Roaming\dione\bin\uv
uv installed successfully
Downloading Visual Studio Build Tools bootstrapper...
Ensured Visual Studio Installer processes are not running.
Launching Visual Studio Build Tools installer (Windows 10 SDK 22621)...
i think its working?
viral mason
has it opened anything?
obtuse summit
nope
viral mason
probably not then
viral mason
<@&1159293204038955078> I need help with helping lol
viral mason
oh nice
viral mason
why is my name yellow now?
leveled up I suppose
obtuse summit
ok cool
99%
almost done
done!
i think its going
ITS DOWNLOADING AGAIN DANG IT
done
Buggy af
imma use the zip
nocturne mural
viral mason
you should prob delete this since it's copyright
obtuse summit
why oh k
viral mason
there are rules against sharing datasets and copyrighted audio here
original server got banned for something like that before
viral mason
it's ok
nocturne mural
ITS DOWNLOADING AGAIN DANG IT
I've seen users having issues with Dione; unfortunately, I can't modify the website to handle that, and besides, the Dione project has just been abandoned
obtuse summit
alr so i should have put applio in english
obtuse summit
I've seen users having issues with Dione; unfortunately, I can't modify the webs...
oh k
nocturne mural
obtuse summit
hol up
viral mason
nocturne mural
craggy brook
help
low shard
nothing is working when i turn it on
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
craggy brook
How can we add a sound we want as an AI cover, or does anyone know how to do it?
viral mason
i need to go to training right?
for making a model yes
viral mason
qhat should i piy
your dataset is around 3 or 4 minutes so batch 4 is fine, train to maybe 300 epochs and test each one with samples starting around 20
low shard
How can we add a voice actor, as in the model?
are you asking how to make AI Covers? what's your pc gpu and os?
obtuse summit
your dataset is around 3 or 4 minutes so batch 4 is fine, train to maybe 300 ep...
alr
random bison
im using Applio for real-time voice conversion but the models I download from sites don't have an index file and I can't generate one in Applio either can anyone help?
wraith tulip
I want to build a local AI pipeline on my own hardware and was wondering if my plan makes technical sense or is realistic. The goal would be to have requirements analyzed automatically and generate structured test cases from them. Do you have any idea how i can do that? thx
viral mason
im using Applio for real-time voice conversion but the models I download from si...
what models are you using? just get them from here https://discord.com/channels/1159260121998827560/1175430844685484042 or from https://voice-models.com
random bison
huggingface
im turkish so my voice is cracking when im using other language models so there is so many turkish voices
viral mason
huggingface is unreliable for finding voicee models unless you just mean the downloads from here
random bison
well are there any real-time voice changer programs that only have the .pth extension?
viral mason
no, they use both but it's not required for index
what is your pc gpu (Nvidia or AMD)
random bison
tyy
viral mason
you're welcome! first one is the voice changer Vonovox, second one is a virtual cable fr using the voices you want on discord or in games
random bison
would VB Cable do the same job?
hardy yew
would VB Cable do the same job?
yeah, in general both are similar, it's just that sometimes one causes issues and the other doesn't. And it seems that for some people VB is more likely to cause troubles
but usually it's just as fine
(if it works - it works)
viral mason
would VB Cable do the same job?
yea for some reason trying to setup the exact setup I have on my pc onto my laptop just causes like the heat death of the universe where the mic slowly just buzzes out of existence but it's completely fine on my main pc
random bison
there is little bit a robotic voice Is there any way to remove it for vonovox
viral mason
it's most likely the model you're using
or your mic could be picking up things happening in the background like a tv or a fan
random bison
i think my mic's good so its probably from model
wise beacon
im trying to get the voice changer to work with my focusrite but nothings working
viral mason
i think my mic's good so its probably from model
did you try raising block size just a little to see if it improves quality? it will cause a little more delay though
random bison
noooo tysm its better nooww
viral mason
noooo tysm its better nooww
You're welcome!
tame oracle
noooo tysm its better nooww
what is that server tag my guy
hallow thistle
tame oracle
is it bad 🙁
Its just so long lmao 😭
viral mason
looks like an ancient curse
hallow thistle
is it bad 🙁
Try switch to something English tag.
kindred fjord
hi, what should I do if the sound is doubled or even tripled when I use vcclient in discord?
inner tide
Where can I get an OpenAI API key
hallow thistle
hi, what should I do if the sound is doubled or even tripled when I use vcclient...
What is your PC GPU? Did you follow any tutorial or guide before? And what do you use the voice changer for?
fiery raptor
why does ts keep happening
full sky
Has anyone downloaded all of the Tekken 8 Characters? For me I only got Hwoarang downloaded ;-;
low shard
why does ts keep happening
what's your PC GPU, OS, and what are you doing exactly?
low shard
hi, what should I do if the sound is doubled or even tripled when I use vcclient...
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
fiery raptor
what's your PC GPU, OS, and what are you doing exactly?
I have 3070 windows 10 n im trying to use the Applio (v3.6.2) RVC (local). For some reason my model's index file keeps bugging out.
low shard
I have 3070 windows 10 n im trying to use the Applio (v3.6.2) RVC (local). For s...
I'm guessing you're trying to do Inference, can you link the model you're trying to use, and try to use another model?
fiery raptor
I'm guessing you're trying to do Inference, can you link the model you're trying...
I did. Other models work but the model i used was a downloaded one from weights.gg before they closed.
I have the file for it, but its not my model so i probably dont have the tools to use it
low shard
I did. Other models work but the model i used was a downloaded one from weights....
can you send the file?
It might be a broken model
low shard
it came a .json, .pth and .index
send the whole zip
also, the json was just metadata, you don't need to use it at all
fiery raptor
low shard
I hope you're not using the .json
low shard
it came a .json, .pth and .index
what if you use only the .pth without the .index?
fiery raptor
what if you use only the .pth without the .index?
i do. I tried asking chatgpt for help and it gave me a few solutions though the error keeps occurring
did i scare u away w the gpt mention
low shard
i do. I tried asking chatgpt for help and it gave me a few solutions though the ...
I just tested the files myself
The issue is a corrupted/empty .index file
Your only solutions are:
- In Advanced Settings, set Search Feature Ratio to 0 so it doesn't use the Index
- Try Another Model
you can find other models in #1175430844685484042
with the first solution, it does work but you won't use the RVC Voice Model trained accent
fiery raptor
I just tested the files myself
The issue is a corrupted/empty .index file
Your...
alright, thanks man 
low shard
yw
fiery raptor
if only weights never shut down
low shard
if only weights never shut down<:cat_doom:1420282253459591240>
There are few people who still use RVC after the hype died, and most people don't want to pay, so that's why
most people liked it for being free and a big archive, but that wasn't going to be sustainable
most services like google and youtube which seems completely free, get paid with your data and ads instead
fiery raptor
most people liked it for being free and a big archive, but that wasn't going to ...
i totally get what you mean. I did use it free as well but it makes sense you can't just run a whole company if no one gives you a dollar
silk meteor
hallow thistle
What is your PC GPU? What do you use the voice changer for? And did you follow any tutorial or guide before?
What the? Are you using the program as E-girl or catfishing or something?
tame oracle
My GPU is AMD then 
viral mason
My GPU is AMD then <:gura_sideeye:1420419744028164126>
use wokada tg fork for voice changer :3
tame oracle
use wokada tg fork for voice changer :3
😭
I am NOT
Using AI on my VEGA 7
IGPU
😭
but thank you
viral mason
what kinda dragon ball character is that
tame oracle
lmao 😭🙏
viral mason
vonovox, it's for Nvidia gpus only
worldly kayak
guys help i talk and i ceant hear it at all it sounds like i got no mic
little ether
Hello guys , i am Prince , from India , I am new to discord and to this server ,can anyone help me with how to using it
worldly kayak
why does it take a long time to talk and wheen i change it to anything lower than 256ms it stops working
verbal birch
How to optimize the gpu for training qwen-3 8b with full dataset roleplay?
grim solstice
does anyone know why my voice changer keeps cutting off when im trying to use it like its very delayed aswell
very very delayed
and im barely able to talk
u cant hear anything im saying
low shard
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
low shard
why does it take a long time to talk and wheen i change it to anything lower tha...
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
low shard
does anyone know why my voice changer keeps cutting off when im trying to use it...
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
Please yall don't ignore the Sapphire Message, Elaborating is important
viral mason
# Please yall don't ignore the Sapphire Message, Elaborating is important
these people... smh
finite wind
hey I stumbled onto a dataset when I tried to gather some, and it turns out it's 96khz WAV
should I downgrade it to 48khz before I use it? I never got this high quality one before
viral mason
should I downgrade it to 48khz before I use it? I never got this high quality on...
you should really train on 32k tbh 40 and 48k is kinda harder to train on
rare stream
im using the onnxdirectML-cuda latest version, i have an amd radeon 6600xt, but even when i set it to use gpu 0, (it is gpu 0 and not anything else i checked in task manager) it still uses my cpu, how do i fix this?
viral mason
im using the onnxdirectML-cuda latest version, i have an amd radeon 6600xt, but ...
since you have amd you should be using Wokada tg fork, whatever you have currently is outdated, if you got it from a yt tutorial
spark prawn
Ive been out of the loop on ai voice for a few months now, but i was looking to get back into things
rare stream
since you have amd you should be using Wokada tg fork, whatever you have current...
can you please tell me where to get the official version from?
viral mason
Ive been out of the loop on ai voice for a few months now, but i was looking to ...
explain what ur lookin into, like training or just using voice changer
viral mason
can you please tell me where to get the official version from?
I'll get you the download
-rt
spark prawn
explain what ur lookin into, like training or just using voice changer
I mostly use it for creating short character videos
So i act out the voice and then use ai to have a range of charecter types
I can show an example of video links are permited
viral mason
I mostly use it for creating short character videos
for doing that you'd need to use Applio
you can use it either locally or on a website like colab or kaggle
viral mason
I can show an example of video links are permited
I think I know what you mean
like those shorts with star wars figures talking to eachother about random stuff right
spark prawn
Kinda? In this case its Vrchat movie shorts (advertising stuff or funny skits)
viral mason
ooh I love using the voice changer in vrchat
spark prawn
It wouldnt be live though, ive noticed the massive lag creates too much of a delay for good communication
low shard
im using the onnxdirectML-cuda latest version, i have an amd radeon 6600xt, but ...
what's your pc os? what are you trying to do, TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay? What tutorial link did you use?
viral mason
you can record what u want and imput it to a model
spark prawn
I heve a trained model from 2 years ago i was using with RVC
spark prawn
you can record what u want and imput it to a model
Ill look into it! Can you use your own models as well?
viral mason
Ill look into it! Can you use your own models as well?
Btw here's the docs for applio so you don't have to look around at outdated stuff on Google
spark prawn
Btw here's the docs for applio so you don't have to look around at outdated stuf...
Appreciate it!
inner tide
Where can I find an OpenAI API key
spark prawn
I sometimes use the live stuff on vrc as well, if im presenting a booth at a convention and have to be in character it helps to not break the imersion
inner tide
I need the API key to make a basic chat bot
low shard
Where can I find an OpenAI API key
inner tide
Thanks
viral mason
I sometimes use the live stuff on vrc as well, if im presenting a booth at a con...
fair yea, if you have Nvidia you should use Vonovox, otherwise if u have AMD use wokada tg fork, they're much more optimized than older stuff from yt
those would be for realtime use
finite wind
you should really train on 32k tbh 40 and 48k is kinda harder to train on
can you elaborate on how 40 and 48k are kinda harder to train on? my only guess is more artifacts for whatever reason
what I gather in the server is that 48k pretrain sucks and worse sibilants and breaths than 32k hmm
cursive cypress
AMD Radeon RX 5600M (Pretty old notebook that somehow still works...for most of the time)/ Windows 10
I don't really like using my mic. So 99% of the time I stay muted when talking with friends on Discord. I ended up stumbling upon the discord /tts and found the voice to be extremely annoying. My friends seemed to agree with me after I spammed 'meow' one too many times. So I thought to myself: is there any good tts programs out there? And I discovered TTS Voice Wizard. And apparently you can use w-okada with it. And that is what I was gonna do... until I saw this in this server: https://discord.com/channels/1159260121998827560/1220262456975298610
To my understanding the owner of the post said it is a TTS only. But I am not well-versed in AI and almost a tech illiterate, so where and how do I use this? Can I use it with okada? Will I get better results with a different program?
coarse lagoon
hello how are you humans I want to discuss about the best Ai models now Idk what is the best Ai's can an expert guide me in this field I would be very thankful
spark prawn
fair yea, if you have Nvidia you should use Vonovox, otherwise if u have AMD use...
Im on a 3080ti, it should be enough i think?
viral mason
Im on a 3080ti, it should be enough i think?
yea that should be enough to run Vonovox
I'll get you the downloads
first link is to vonovox, and second one is a virtual cable to connect the voice changer to vrchat or discord ect
spark prawn
first link is to vonovox, and second one is a virtual cable to connect the voice...
Thanka again! I was using a combination of rvc and bannanameter last time
wise beacon
How do I get vonovox to work with a focusrite
viral mason
How do I get vonovox to work with a focusrite
what is that?
it is only avaialble for Nvidia gpus so if you have something else it won't work
wise beacon
it is only avaialble for Nvidia gpus so if you have something else it won't work
i have a nvidia gpu its a audio interface for my mic
viral mason
i have a nvidia gpu its a audio interface for my mic
oh
chrome ember
i am trying to train a model using Mangio-RVC v23.7.0. it does all the stages fine until the last stage where i click "Train model", then it just stops with some asian text followed by train.log . Could it be that the RTX 5090 isn't properly supported in pytorch?
because i remember this being a problem with other things running with pytorch. I am getting this error among other things in the console:
viral mason
i am trying to train a model using Mangio-RVC v23.7.0. it does all the stages fi...
mangio rvc is extremely old, have you tried applio?
chrome ember
mangio rvc is extremely old, have you tried applio?
not yet. though it was another option I had in mind
viral mason
are you using it locally btw?
I can't help much with it if so but I can get you the docs which should have the download
-rvc
chrome ember
yes, locally. I will do as you suggested and try applio. If it is better maintained, it should hopefully have a newer version of pytorch that supports blackwell GPUs
cursive swift
why is mine so laggy?
chrome ember
@viral mason got applio running and it seems to be doing stuff now, will be interesting to see the result. thanks 😄
low shard
why is mine so laggy?
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
viral mason
<@1023278814752677918> got applio running and it seems to be doing stuff now, wi...
you're welcome!
cursive swift
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc o...
idc what my gpu is this is all so confusing
chrome ember
hard to help without any context
cursive swift
idc what my gpu is this is all so confusing
adm radeon rx 6500 xt gpu. im trying to use a voice changer
low shard
adm radeon rx 6500 xt gpu. im trying to use a voice changer
you need to still elaborate:
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
there are over 10 apps you could call voice changers, this isn't only a voice changer server, this is a general ai discord server, if you don't elaborate we can't really help
Those aren't 1 click things such as using chatgpt
cursive swift
you need to still elaborate:
- your pc os
- what you're trying to do, like TTS, ...
MMVCServerSIO is what the file is called
low shard
MMVCServerSIO is what the file is called
please elaborate everything, that could mean different programs and versions
viral mason
you're using an old program, what is your pc gpu (Nvidia or AMD)
also what are u using it for just curious
wide perch
Whenever I select "Client" in RVC instead of Server, it says
"Warning Select an audio input device."
Even though I have my mic selected in the web.
wintry locust
Hey whats a good voice changer for someone like a rappper, I wanna make a remix just for a little self project
viral mason
you could just use applio, you record the audio then put it over a model of a rapper
doesn't need to be voice changer since those are for realtime
-rvc
ashen bolt
Your PC GPU of GTX 1050 & Your PC OS is Windows 10
grand plinth
how can i train/make voice models?
grand plinth
never mind
lyric nebula
Yo so I tried to train my own AI voice on applio but it was a disaster even that I provided him 50 minutes of the voice, so I was wonderingw, is 50 enough and how do u do to don't make it sound robotic or Muffled ?
And I got an 3070
hardy yew
Yo so I tried to train my own AI voice on applio but it was a disaster even that...
There's a lot of factors and variables affecting the final model.
Sample rate matters. 32k models are the most forgiving and usually better.
Data -> length, diversity, quality, also the voice properties matter and can be handled better/worse
Preprocessing obviously matters too, if present.
Then with training, pretrain matters (though OG is a good pick), so does batch size.
And then, various checkpoints might be vastly different so you should pick the best epoch by ear
(but to answer your question: 50min is more than enough to make a nice sounding model)
lyric nebula
There's a lot of factors and variables affecting the final model.
Sample rate ma...
But with the actual state of the technology can we reach an real time voice who don't sound muffled or weird ?
hardy yew
But with the actual state of the technology can we reach an real time voice who ...
Yeah, great sounding voices are possible.
Definitely not "perfect" and with limited flexibility, but possible
lyric nebula
Yeah, great sounding voices are possible.
Definitely not "perfect" and with limi...
Normal conv for gaming ?
I also heard about voice.ai
But idk
hardy yew
Normal conv for gaming ?
Sure
Biggest issues in RVC are non-verbal sounds
Laughter, grunts, screaming etc
But expressive speech in general can be quite nice and realistic
lyric nebula
Laughter, grunts, screaming etc
Then I will just act calm, do u have a model to advice me ? Mine sound like a grimlins 😭
hardy yew
There's a lot of free models available in #1175430844685484042 and on https://voice-models.com/
Lots of them are gonna be bad, but there's definitely some nice ones too
I don't know any particular ones, I don't really use realtime much
Mostly just enjoying discussions, supporting people and playing around with my trainings
In #1175430844685484042 you always have voice samples so you can listen to various models easily
Should give you a nice glimpse into what RVC is capable of
(my Anran model, for example, has lots of samples - male/female singing as well as casual speech and even poorly converted non-verbal sounds - you can hear strengths and weaknesses that way)
lyric nebula
There's a lot of free models available in <#1175430844685484042> and on https://...
U use applio ?
hardy yew
More or less. I use Codename's fork, but it's almost the same thing.
Entire training procedure is the same
Slight differences in preprocessing, but that step is mostly user-based anyway
hasty junco
can i anyone help me with voicechanger
slow aspen
here's the basic audio separation guide, just do this for all audio you're using...
Can I ask why did you choose HyperACE v2 Vocals instead of ver 2025.07?
hallow thistle
can i anyone help me with voicechanger
What is your PC GPU? And what do you use W-Okada for?
void raft
Is there any other repository with AI voices? Since weights closed down I have lost access to a lot of voices I need for work, this server's repository is sadly does not have a lot of voices
hallow thistle
Is there any other repository with AI voices? Since weights closed down I have l...
hardy yew
low shard
Whenever I select "Client" in RVC instead of Server, it says
"Warning Select an...
Please elaborate more your PC GPU, os, what you're trying to do and the tutorial link
RVC means Retrieval-based-Voice-Conversion not realtime voice changer
low shard
How to get voice models?
void raft
@hardy yew @hallow thistle Sorry but no, both of these have no voices I need 🙁
low shard
<@214134626065580033> <@561807329977434122> Sorry but no, both of these have no ...
Then #1159289738314919936 or make it yourself
low shard
Is there any other repository with AI voices? Since weights closed down I have l...
Most of weights model were from #1175430844685484042 , but most models made on weights or uploaded only on weights are lost forever
Making an archive would take way too much storage and money
Your last option is checking https://HuggingFace.co/models
hardy yew
<@214134626065580033> <@561807329977434122> Sorry but no, both of these have no ...
As Nick said, you can request a model (maybe someone will be eager to make it for you) or make it yourself.
Also there's some places (I've heard of booth, kofi? Perhaps there's much more) where models can be bought. I guess some of those should be much better than the free ones. But personally I can't think of a scenario in which I'd want to buy an RVC model
void raft
As Nick said, you can request a model (maybe someone will be eager to make it fo...
Well, no better time to learn as now I guess... thank you
void raft
As Nick said, you can request a model (maybe someone will be eager to make it fo...
Will 4090 be enough to train good RVC or it is more cpu dependant?
hardy yew
Will 4090 be enough to train good RVC or it is more cpu dependant?
Training uses GPU, but a 4090 is decent for sure
Honestly any RTX 20x0-50x0 will be good, it's just that better cards make the job much quicker (also larger VRAM means you can use higher batch size, but you probably won't need to do that)
I was satisfied with my 4070's performance before I moved on to training in cloud
So 4090 will be nice for sure!
For me training in cloud is more convenient. Less comfort compared to running it all on your own PC, but doesn't need you to leave the PC running or force you away from gaming for a couple hours xD
low shard
For me training in cloud is more convenient. Less comfort compared to running it...
until free limits hit you
hardy yew
(just in case, cloud training is a paid service in general, but there are places with limited free usage. I use kaggle which grants 30h of free GPU training weekly)
hardy yew
until free limits hit you
Yeah, was just about to say it xd
dim vigil
Hey is anyone living in the US and interested in developing a new concept in AI? I have a working prototype that is using the resources of a 14B LLM but the knowledge of a 72B General, a 32B coder, a 14B Coder, a 7B coder, and a 70B medical all rolled into one with no fine tuning and the 14B speeds. There is a lot more to the project that I'll discuss in private. If so hit me up in a DM. I don't need help, looking for someone with the same ideals open to new concepts.
dim vigil
I'll be in voice if anyone want to chat or has any questions. Feel free to join me.
viscid blaze
Hey, I've seen the Wokada Voice Changer all over YouTube. Is there a better one?
At some point on this discord, I already had this conversation and got told that this one is outdated and not really "good" anymore.
I think Vonovox is a good choice, right? For Realtime.
hallow thistle
Prepare yourself for the bad thing will happen.
viscid blaze
Oh, ur gone bro.
Read the rules
E-Girl trolling isn't allowed / supported and will be punished
craggy brook
What I want: Can you create a voiceover for Yuuko Sumitomo, who voiced Millia in the GGST game? I'd like to sing a song with her voice (I actually messaged her directly on Twitter, but it's unclear if she responded).,
patent trellisBOT
hallow thistle
What I want: Can you create a voiceover for Yuuko Sumitomo, who voiced Millia in...
Do you think messaging a voice actress on X/Twitter as an individual would work? Only handful people "in the industry" can make that possible.
viral mason
Can I ask why did you choose HyperACE v2 Vocals instead of ver 2025.07?
Is the one you said better than hyper ace?
low shard
Hey, I've seen the Wokada Voice Changer all over YouTube. Is there a better one?...
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc os
- what you're trying to do, like TTS, E Girl Trolling / Catfishing, AI Covers, Training RVC Models, Roleplay
- the tutorial link used
craggy brook
Do you think messaging a voice actress on X/Twitter as an individual would work?...
I don't know how it happens, but celebrities can do it. Some get the message by chance.
viscid blaze
this is a general ai discord server, please elaborate:
- your pc gpu
- your pc o...
My fault
- 4060 TI with 16GB
- A custom OS (FSOS)
- Roleplay with a friend
- No tutorial used. Yet I've only installed Vonovox, tried to tweak the settings and search a model.
low shard
My fault
- 4060 TI with 16GB
- A custom OS (FSOS)
- Roleplay with a friend
- No ...
FSOS? Never heard about it before
Is it a Linux distro?
viscid blaze
Windows but debloated and tweaked. For competitive gaming.
slow aspen
Is the one you said better than hyper ace?
Idk last time I used any models was over the year ago, so Im behind in knowledge which is the best now. But higher SDR = better right?
low shard
Windows but debloated and tweaked. For competitive gaming.
Eh don't know how much to trust those, be careful especially if you only saw it in a random YouTube tutorial, might be better to debloat it yourself
-realtime
You could try Vonovox or wokada tg develop fork
hallow thistle
I don't know how it happens, but celebrities can do it. Some get the message by ...
I'm quite skeptical because this strategy sounds unprofessional. To actually contact the voice actress for a collaboration, you would contract through related agencies and companies. Are you aware that not every project would be free of charge, right? 
low shard
Not sure if your windows fsos might cause issues tho
viscid blaze
Eh don't know how much to trust those, be careful especially if you only saw it ...
It's an popular OS with a discord over 50k members, had their optimizations multiple times, they are pretty trustful.
craggy brook
I'm quite skeptical because this strategy sounds unprofessional. To actually con...
yes
viscid blaze
You could try Vonovox or wokada tg develop fork
I'm trying Vonovox but I'm having issues with it sounding too bad to actual believe.
Not believe but more likely, even sound anything near a human
low shard
It's an popular OS with a discord over 50k members, had their optimizations mult...
Your choice
low shard
I'm trying Vonovox but I'm having issues with it sounding *too bad* to actual be...
What RVC model are you using? Did you try others too?
viscid blaze
I tried these:
https://ptb.discord.com/channels/1159260121998827560/1327262488277815356
https://ptb.discord.com/channels/1159260121998827560/1198491983442878545
I have a kind of deep voice so I didn't find any good model for my voice yet.
Might be the fact, that it's a singing RVC model and I'm literally only talking but there is one more that I forgot, which also didn't work so overall I tried 2 yet.
I will try some models and get back to this chat later.
viral mason
Idk last time I used any models was over the year ago, so Im behind in knowledge...
I would think so but I'm not entirely sure, I use whatever gives the best results from my experience with keeping the voices as clean as possible and removes any sound effects like water or cartoony effects ect
viral mason
I tried these:
https://ptb.discord.com/channels/1159260121998827560/132726248827...
Have you tried raising the pitch
viscid blaze
Have you tried raising the pitch
Yes, I just sound robotic
viral mason
Hm
viscid blaze
I'm trying the tg develop fork right now.
viral mason
Try turning the block size up to 50 whenever you switch back to Vonovox
viral mason
I'm trying the tg develop fork right now.
Turn crossfade up to max and for each model turn protect down to 0 in the advanced settings, it makes models sound better
I've tried it and also a friend has said it works well for him
viscid blaze
Thank you, I'll try that.
viral mason
👍
supple trail
is a rtx 5070 ti 16 gb good enough for local?
viral mason
I'd lower chunk size until the delay is good for you
supple trail
using fcpe onnx
viral mason
is a rtx 5070 ti 16 gb good enough for local?
Definitely yea!
Use rmvpe not onnx
Less quality
supple trail
fcpe onnx tho.
viral mason
On fcpe
supple trail
whats the quality difference tho
viral mason
Test it trust
supple trail
i did
viral mason
Oh
viral mason
Btw why onnx, if you're on amd it would make sense
supple trail
onnx faster tho?
viral mason
Is it?
supple trail
yeah
viral mason
Ah
supple trail
ONNX models can run 4–5x faster than their PyTorch counterpart
viral mason
Vonovox doesn't have the onnx models of either of those from what I checked so I can't really test
supple trail
for rvc wokada
viral mason
Current beta is super quick and has better quality from what I've heard
I use it too
viral mason
but im more used to kaggle since yeah i pretty much code stuff there
I use Kaggle only for model training since I like having space on my pc
supple trail
do you think its possible to achieve perfect replication of "sighs, screams, humming, etc" ?
viral mason
Best bet is using the legacy core 1.5 pretrain
I've been using it for a while it's pretty peak
viscid blaze
Is there a good female voice you could recommend?
viral mason
No, is there a reason you want a female model in particular?
viscid blaze
I want to roleplay with a friend and am playing a female character.
viscid blaze
Alright, thank you!
viral mason
I don't use regular human voices often unless they're a specific character I like from a series or game so I wouldn't know the best to choose
viral mason
do you think its possible to achieve perfect replication of "sighs, screams, hum...
You use applio to train right?
viral mason
Fair enough
Making your own models will pretty much always get the best results especially if the character or person you want has a really old model that is kinda bad or just no model at all
supple trail
tbh i have a feeling in the future everything is gonna be way easier
supple trail
Making your own models will pretty much always get the best results especially i...
yep
viral mason
It's what I do and have done specifically just because I enjoy using the voice changer
supple trail
btw i implemented hrzn support to the kaggle tg develop wokada fork and created my own uploading method + pulling method that downloads all the models into the model_dir
viral mason
tbh i have a feeling in the future everything is gonna be way easier
Rvc won't have any breakthroughs until some rich smart guy picks up where rvc boss abandoned it all those years ago
Since nobody knows how the original pretrain was made nobody can recreate it or beat it
Only thing I'd say is better than it is legacy core v1.5
And that's just a fine-tune of the original
Not a true pretrain
supple trail
btw fcpe is prob better than rmvpe in some things
non-verbel support + stability + latency + best use case : expressive performance, singing, and emotional reactions + accuracy : higher precision, tracks fast, downside: unstable pitch.
supple trail
fcpe yh my bad
viral mason
Oh lol
supple trail
typed too fast
viral mason
I do that all the time
supple trail
when you switch f0_method to fcpe its really impactful:
captures the human element more better cuz unlike RMVPE which tries to force your voice into a clean curve
and handles the energy quite good in a scream
downside: requires a chunk increase around 128-160
viral mason
downside: requires a chunk increase around 128-160
Psst
magic hare
Hey, what can I do with this to make my voice more clear and more easy to understand, and I guess also just be faster? I'm noticing a good amount of delay when I speak.
low shard
Hey, what can I do with this to make my voice more clear and more easy to unders...
did you come from a youtube tutorial?
magic hare
did you come from a youtube tutorial?
a while ago, yeah. Haven't changed my settings in a bit.
low shard
a while ago, yeah. Haven't changed my settings in a bit.
those are all old, you seem to have gotten an outdated version
uninstall it and uninstall vb audio cable
magic hare
What should I be using, then?
low shard
What should I be using, then?
you seem to have an Nvidia RTX 3060 GPU on Windows 11 right?
then you can try either vonovox or wokada tg develop
-realtime
magic hare
you seem to have an Nvidia RTX 3060 GPU on Windows 11 right?
Windows 10, but yeah
low shard
Windows 10, but yeah
should still work fine
magic hare
should still work fine
Which one should I download? I don't use an AMD.
low shard
-realtime
low shard
Which one should I download? I don't use an AMD.
please read the guide, everything is in the guide
you should get the 2 windows cuda ones
AMD64 means the CPU architecture, all modern 64 bit CPUs use the 64 bit extension made by AMD
viral mason
Which one should I download? I don't use an AMD.
if you have Nvidia use Vonovox
low shard
if you have Nvidia use Vonovox
they can try both
viral mason
true
low shard
they can just check the pros and cons of each
viral mason
just recommending the best first
magic hare
if you have Nvidia use Vonovox
I just grabbed wokada, but ig I'll try that one instead
viral mason
I just grabbed wokada, but ig I'll try that one instead
I'll get the download link for you
the beta is better than the current release
magic hare
👍
viral mason
do you have vac lite btw? it's a virtual audio cable that you'll need top use either voice changer software in discord or any game
viral mason
is it vb cable?
magic hare
I think so
viral mason
that one isn't recommended as much due to it causing odd issues on windows sometimes
magic hare
I should use a different one, then?
viral mason
yah, this one should work fine with no issue ^^
https://software.muzychenko.net/freeware/vac470lite.zip