i dont know what to get anymore
#✨│ai-help
fiery topaz
fiery topaz
is it better
viral mason
first one is the voice changer second one connects it to games or discord so ppl can hear it
viral mason
is it better
it's the current best
fiery topaz
k ty
ebon tapir
first one is the voice changer second one connects it to games or discord so ppl...
ik but when it installs which oe do i run
theres like bunch off things on it
viral mason
first extract them both then, for vac lite run setup 64 and then install driver
then for vonovox run setup
viral mason
literally the easiest thing ever
ebon tapir
viral mason
also are u sleep deprived possibly, you're making a lot of typos lol
ebon tapir
ima be honest i dont know what u said there
viral mason
are you ok?
ebon tapir
its okay dw
viral mason
first extract them both then, for vac lite run setup 64 and then install driver
for the second download run the file called setup64, then after that click install driver
and for the first one run the file called setup
ebon tapir
oki ty
ebon tapir
this is taking ages to extract
viral mason
this is taking ages to extract<:cat_bruh:1167393213661839433>
most likely that's due to wifi
ebon tapir
most likely that's due to wifi
awww
no worryes tho everything to be soldeir boy
can i ask a question tho
viral mason
sure
ebon tapir
so when i used a voice model on okada it was like glitching will it glitch w this ne???
one
viral mason
so when i used a voice model on okada it was like glitching will it glitch w thi...
I'm not sure, it shouldn't tho
ebon tapir
I'm not sure, it shouldn't tho
i might have used bad settingsfor it tho im not sure
viral mason
i might have used bad settingsfor it tho im not sure
I just realized I told you the wrong file to run for vonovox, it's not "setup", it is "start"
sorry
dawn copper
It's all fixed now! After switching to Wokada TG and Vac Lite, I was still having the issue. Buut the new UI is way better than the one I used before!
Anyway, for anyone that has the same problem down the line... It was my own ignorance in using the software! Don't choose "Client", apparently you want to stay on "Server" mode.
I assumed Server would be like.. Hosting the service for others to use! And Client was, use it like a client side user. I assumed wrong it seems!
And thank you Local Worm, for suggesting Wokada TG! It has a lil graph for the latency, and is a bit more intuitive on narrowing down the settings.
ebon tapir
I just realized I told you the wrong file to run for vonovox, it's not "setup", ...
its okay but this is still 0 percent
the extract is broken..
ebon tapir
that's odd
It got fixed
Btw is there some like good settings
Or js random
Setitings
viral mason
Btw is there some like good settings
have block size between 0.30 - 0.50
fiery topaz
if u need any help just lkm
so the voicechanger changed to this?
viral mason
so the voicechanger changed to this?
nah it's just the newest one and only one that's still actively recieving updates
fiery topaz
k
fiery topaz
nah it's just the newest one and only one that's still actively recieving update...
how do i use this thing
viral mason
how do i use this thing
what do u mean?
fiery topaz
nah it's just the newest one and only one that's still actively recieving update...
how do i make it work
so other people can hear it
viral mason
have you installed vac lite?
fiery topaz
i have cable
viral mason
your settings should look like this
fiery topaz
it does
fiery topaz
its working but how do i make other people hear it
viral mason
its working but how do i make other people hear it
you pressed start yes?
fiery topaz
yup
empty canopy
i want to use a voice trainer but i don't know/can't find any ones that work or aren't too confusing
viral mason
what gpu do you have (Nvidia or AMD) and where did you download the voice changer from?
viral mason
i want to use a voice trainer but i don't know/can't find any ones that work or ...
use applio, I have a video of how to use it on Kaggle
that's super old
u should use Vonovox, what are u using the voice changer for btw
fiery topaz
you pressed start yes?
no im saying like in discord or any other game how do i make it work in there
empty canopy
use applio, I have a video of how to use it on Kaggle
can you send it to me 
fiery topaz
u should use Vonovox, what are u using the voice changer for btw
no im saying like in discord or any other game how do i make it work in there
fiery topaz
u should use Vonovox, what are u using the voice changer for btw
uhm can u answer my question plss?
viral mason
no im saying like in discord or any other game how do i make it work in there
oh, have your mic as line 1 and your output as your headset/headphones
sorry I spilled something give me time
fiery topaz
oh, have your mic as line 1 and your output as your headset/headphones
whats line 1?
viral mason
whats line 1?
the name of the virtual cable when selecting it in your audio settings
fiery topaz
my virtual cable only has output in input for some reason
fiery topaz
the name of the virtual cable when selecting it in your audio settings
my virtual cable only has output in input for some reason
obtuse cedar
how do i fix the res being at like 16000 ms
viral mason
my virtual cable only has output in input for some reason
whym?
empty canopy
r u able to send the applio video?
viral mason
the boxes are the model slots
empty canopy
ty
viral mason
i dont think u understand me
I don't think I do, are you able to send a screenhot?
fiery topaz
ok i got it is there a way that i could hear my self using the voice changer?
empty canopy
what do the sample rates mean on vonovox
viral mason
what do the sample rates mean on vonovox
it's an upscaler, if you set it to say 48000 it will upscale the model to that sample rate
obtuse cedar
it's an upscaler, if you set it to say 48000 it will upscale the model to that s...
i have the voice changer and the res keeps going way up and it makes the voice like get super delayed and laggy
viral mason
i have the voice changer and the res keeps going way up and it makes the voice ...
which one are you using?
viral mason
there are 3 main ones that are used, Vonovox, Wokada tg fork and, wokada deiteris
obtuse cedar
there are 3 main ones that are used, Vonovox, Wokada tg fork and, wokada deiteri...
can you give me one that works for AMD
because idk if i downloaded the right one
obtuse cedar
sure! 1 sec
what do i use to start the voicechanger cuz i installed it
viral mason
what do i use to start the voicechanger cuz i installed it
after u extract it run mmvcserversio
obtuse cedar
after u extract it run mmvcserversio
okay
viral mason
u have a virtual audio cable right? like vb cable or the one I sent?
obtuse cedar
u have a virtual audio cable right? like vb cable or the one I sent?
yeah i downloaded the one u sent
what do i do now that it says this
2026-04-16 21:14:35,853 INFO [MMVC_SocketIOApp] Initializing...
2026-04-16 21:14:35,858 INFO [MMVC_SocketIOApp] Initialized.
2026-04-16 21:14:35,946 INFO [server] --------
2026-04-16 21:14:35,948 INFO [server] The server is listening on http://127.0.0.1:18888/
2026-04-16 21:14:35,952 INFO [server] --------
empty canopy
it's an upscaler, if you set it to say 48000 it will upscale the model to that s...
and the embedder?
viral mason
what do i do now that it says this
2026-04-16 21:14:35,853 INFO [MMVC_Socke...
oh nvm this is is normal my bad
did it open?
viral mason
what do i do now
set your mic settings like this
input will be your regular mic, output is your virtual mic
obtuse cedar
set your mic settings like this
okay
hallow thistle
what do the sample rates mean on vonovox
Basically audio stuff used in most programs, but not sure which part to explain.
viral mason
<@&1159293140440723499> ?
hallow thistle
Well, I can't help with either catfishing, catching a p or E-girl/E-boy model, because you know what these are prohibited in this server.
empty canopy
Basically audio stuff used in most programs, but not sure which part to explain.
lower vs higher
wide oasis
trying to be funny and then saying catching preds is suspicious
hallow thistle
lower vs higher
What do you mean by this question?
wide oasis
Why trolling?
...
...
are you saying that your friend is a pred
well as a wise man said once you are the ppl you associate yourself with
i mean you already caught him ath this rate since you know hes a pred and indian
like scamming him
sspecious
its not your job call the indian police
empty canopy
What do you mean by this question?
whats the differences by higher sample rate vs lower
wide oasis
then explain this
you knew he was indian and a pred
soo you already threw yourself as an offering to preds
are you into that shit
is that a secret words to catch preds too
wait how do you know how they speak if your not one of them
i am not a bro of a pred sorry
@hallow thistle call the mods let get him banned
hallow thistle
whats the differences by higher sample rate vs lower
Higher sample rate (like 48000 Hz) means higher audio quality. For most people, between 44100, 48000 Hz or more is negligible, but any sample rate below 32000 Hz can give lower audio quality.
hallow thistle
empty canopy
Higher sample rate (like 48000 Hz) means higher audio quality. For most people, ...
okay! thank you but one more question on the vonovox, what do the block size + extra time meaning
wide oasis
long story short your obviously trying to do something shady soo we refuse to help we can tell from your smell
hallow thistle
okay! thank you but one more question on the vonovox, what do the block size + e...
One more question? For more information about Vonovox, there's a guide doc. https://docs.aihub.gg/realtime-voice-changer/local/vonovox/
empty canopy
okok tyvm i mustve mised that
half tendon
anyone know how i can take text / text's and derive the general topic from those texts? i have tried things like searching for the most common word in said texts but that didnt work to well. currently the text is chunked and embedded in a local database that a model would need to pull from.
tried using a summarization model but newer transformers versions arnt taking " summarization " in the task field
fiery juniper
i wanna like, train my own voice model that gives me its own index and uhh yk
so i can insert it to wokada or something
IS THEREEEE A WAY TO DO THISS? im kinda newww
should i use replay or applio
gdhdashds
idkk
finite wind
I have this 96khz audio data here but spectrum wise it doesn't go up to 48k, almost like 21.5k on the chart so it's 43k range really
should I be concerned about those silent areas on the top and downgrade to 32k audio to ensure those silent parts are removed or should I just keep it as what it is?
finite wind
because you know, empty spaces are automatically let RVC to make whatever random noises it can fit in hence possible artifacts
hardy yew
This kinda depends on what sample rate you're going to train with. But assuming it's 32k, you can just downsample to 32k, as you're going to want to do that anyway
The main point is to not train a model with sample rate higher than your input. So if it peaks at 21.5k, don't train 48k, you can choose 32k or 40k
hardy yew
Like you said, empty space in the spectrum will cause model to hallucinate in that area
hardy yew
Yeah, but the sample rates for training are standardized to 32/40/48k (also that's what pretrains are prepared for) so 41.1 is not among these options
finite wind
it won't let me convert to 40k the adobe audition :c
hardy yew
(usually 32k models are trained BTW as supposedly it's more forgiving and produces better breath noises. But I haven't done any experiments in this direction so all options are good to consider)
hardy yew
it won't let me convert to 40k the adobe audition :c
Oh, it doesn't let manually setting the target sample rate?
That's weird and quite unexpected from that kind of software
If you're sure that's the case then you can export it at a higher sample rate anyway and resample it later with ffmpeg, librosa or something similar
finite wind
Oh, it doesn't let manually setting the target sample rate?
yeah,,,,
hardy yew
(Not sure if Applio resamples automatically the dataset if it's in wrong sample rate - if it does then it's a non-issue anyway, but this would need to be verified in code)
hardy yew
yeah,,,,
Curious, can you not just type in that field? 🤔
(also that's 44.1 BTW, which is there because it's a standard format, used with CDs too)
finite wind
yeah I just checked and there was no 44.1
I assumed there could be one to write it myself but no
surprisingly
oh nvm I found it
ebon tapir
u have a virtual audio cable right? like vb cable or the one I sent?
any good settings??
fierce bone
Maybe someone can shed some light:
I'm currently trying to get Applio v3.6.2 running on a system.
System is currently running Windows 11.
CPU: 5600x / GPU: RX 9070XT
- I'm following this guide: https://docs.aihub.gg/rvc/local/applio/ / Section: AMD on Windows (Precompiled Fix)
The guide says:
Download a compiled version of Applio v3.5.0 or newer from the Hugging Face repo, and unzip it.
- V3.6.2 is downloaded and unzipped to C:\
Download and install the latest stable HIP SDK from the AMD ROCm Hub.
Important: Install components but exclude/deselect the video driver at the bottom of the installer list.
- Done too, but ended up downloading two different versions, and I'll explain that below.
Add the bin folder of your installed HIP SDK to your System Environment Variables (Path): C:\Program Files\AMD\ROCm<YOUR_VERSION>\bin
- Done for two different versions. 7.1 and 6.4.2
Open a command line (CMD) inside the Applio folder and run:
env\python -m pip uninstall torch torchvision torchaudio
env\python -m pip install torch torchvision torchaudio --upgrade --index-url https://download.pytorch.org/whl/cu118
- It won't do anything, so I excluded the "env" from both lines. Yes, I did run CMD from inside the Applio folder.
Download the patch file corresponding to your installed HIP SDK version from the Applio Assets repo and run-applio-amd.bat.
- This is where I got a bit confused, as the latest patch said:
"zluda patcher for hip sdk 6.4.2"
So I therefore downloaded the HIP SDK for "Windows 10 & 11" which is ROCm Version 6.4.2,
assuming that it was important for the patch bat to work,
but I already grabbed the latest for ROCm 7.1.1 and installed. Now both HIP SDK versions are installed, and everything is added to PATH's accordingly.
Edit the file located at rvc/lib/zluda.py. Replace the content with the following:
import torch
if torch.cuda.is_available() and torch.cuda.get_device_name().endswith("[ZLUDA]"):
# disabling unsupported cudnn
torch.backends.cudnn.enabled = False
torch.backends.cuda.enable_flash_sdp(False)
torch.backends.cuda.enable_math_sdp(True)
torch.backends.cuda.enable_mem_efficient_sdp(False)
- I assumed that the guide said "remove everything inside "zluda.py" and insert this piece of text instead. Ironically I inspected the file beforehand, and all those lines were already in there, therefore my assumption.
run your downloaded patch script, then run "run-applio-amd"
patch script. Assuming this looks fine.
I went ahead and ran the downloaded "run-applio-amd" bat file after. However, this gave me a Traceback:
- I am clueless at this point.
low shard
that an over year old original version
what's your pc os? what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay?
low shard
how do i fix the res being at like 16000 ms
This is a General AI Discord Server and there are many voice changers, elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
low shard
how can i cancel this
this is a general ai discord server, what app are you even talking about?
hasty mesa
Hello guys is this computer good for runnong local ia ?
https://pcgamingbcn.com/pc-gaming-amd-ryzen-7-5700x-32gb-1tb-ssd-rtx-4080-super-16gb/
proven pecan
Beta version of Vonovox or release version?
low shard
Beta version of Vonovox or release version?
the beta has latest freshing edge updates with the cost of it might being more unstable
simple ore
- I am clueless at this point.
you did not add HIP SDK/bin folder to the existing Path variable
not some other random variable, not a new variable
finite wind
@hardy yew I'm trying out Smartcutter now and I have a question about its automatic silence adding feature
low shard
please elaborate, there are multiple programs:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
finite wind
<@214134626065580033> I'm trying out Smartcutter now and I have a question abou...
it's adding 100ms ish silences in between some parts and I was wondering
wouldn't those 100ms of silences are potentially create artifacts? Why add those 100ms of silences?
fierce bone
you did not add HIP SDK/bin folder to the existing Path variable
To begin with, I added a new PATH to the system. It didn't explain that I had to replace anything. Even if i do that, it still reports the same.
fierce bone
new entry under that
Please pardon me, but I don't think I follow you. Example?
simple ore
hardy yew
wouldn't those 100ms of silences are potentially create artifacts? Why add those...
The way I understand it, it's supposed to "standardize" breaks between (usually) words to ~100ms. With the main purpose being probably clear separation of features so that ending of one word doesn't get mixed with beginning of another.
Does it insert silences in some unexpected places in your case? It's a machine learning model so its effectiveness probably varies depending on input data. I can imagine it sometimes doing random undesired stuff
finite wind
oh, so there are reasons to add those 100ms but I got it confused because the core principle of silence in the training model process is like
why you would add silences it creates artifacts
given the fact that if it were to be trained with it creates artifacts
but if it's serving as divider of sort then I guess yeahg
hardy yew
Hmmm that's my understanding of it but I don't even completely trust myself with it so probably neither should you
See DMs
fierce bone
First and foremost: Thank you. Second: Running "run-applio-amd.bat" now, gives me a different output: HIP Library Path: C:\Windows\SYSTEM32\amdhip64_7.dll Press any key to continue . . .
Window closes if I press anything. (obviously)
simple ore
amdhip64_7.dll - is it not in HIP SDK/bin?
there may be _6 dll, just make a copy and rename it to _7
fierce bone
I'll check
simple ore
or _6 can be in Windows\system32
fierce bone
It's actually in the bin folder. Made a copy and renamed it returned with the same output. Both files are ironically, also in system32. (Yes, they were in system32 already) Might've fixed it by removing and reinstalling SDK aswell as starting fresh with Applio. Followed steps one more time, and its now running fine. Apparently I could even use the "env\python" strings now which is odd. Seems to've done the trick. So now it launches, with AMD patches applied and using the run-applio-AMD bat script.
hushed sinew
guys where to download MMVC for amd gpu?
low shard
guys where to download MMVC for amd gpu?
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
onyx ivy
is there any way to not sound robotic, and have a delay of 1 second or under with RTX 4070 SUPER? Windows 11, i9-9900KF
hollow haven
whats epokhs?
low shard
whats epokhs?
epochs are a unit of measuring the training cycles of the AI model
they don't mean how good is the model, it's just an info provided on how they trained the model by the model maker
More ≠ better
Less ≠ better
simple ore
First and foremost: Thank you. Second: Running "run-applio-amd.bat" now, gives m...
hold on
so both _6 and _7.dll are availabe, no when it says the path of DLL it is not an error
the error is whatever crashes later
you need to make sure you did the torch uninstall/install lines and it did install properly, and that you ran a proper patch zluda 64.bat
fierce bone
the error is whatever crashes later
I edited my reply as you were writing
Exactly what I just did too, and it seems to be running accordingly now.
simple ore
so all good?
fierce bone
One thing I gotta figure out is why my GPU isn't being used.
simple ore
what does it show under Training tab under 'Advanced Settings' at the top
fierce bone
simple ore
looks good
why do you say it is not used?
when you run inference, did it shows 'compiling in progress' ?
with 9070xt you did not need to use zluda
fierce bone
I supposed i expected it to "WHINE" when used,
its silent as heck, i dont see utilization being raised
simple ore
in task manager you open the performance tab, click on the gpu on left /bottom, there's a chart for VRAM used
if it goes up when it is being used
viral mason
any good settings??
For Vonovox just use block size at the default 0.30 or go up to 0.50 if you want
And for pitching just change that depending on if you're using a female model of a male model
If you're a guy using a female model use pitches 3-12
If you're a guy using a guy model just put pitch at 0
Unless it's one with a high voice like Mickey Mouse
fierce bone
if it goes up when it is being used
All i see is that once I'm trying to convert a 3 second test vocal with any model i've downloaded, VRAM goes up but stays consistent. But literally still after 800 seconds, its not done processing. Seems odd.
slow aspen
Guys did resampling audio to 32k and removing DC Offset make quality worse?
viral mason
dc offset?
hardy yew
where did that data come from?
viral mason
I'm not sure what that is, I usually just truncate the silence in my datasets after I am finished cleaning them
slow aspen
where did that data come from?
wdym?
hardy yew
what's the source of this dataset
DC offset shouldn't really be a thing in most cases, I think
slow aspen
what's the source of this dataset
Ahh its only one song. Im making dataset. I remember there was talk about dc offset while ago and that it can worsen ai separation so i tried removing it before seperating voicals
slow aspen
what's the source of this dataset
Why asking? Does it look bad?
hardy yew
Why asking? Does it look bad?
hard to tell how it looks, as it's too squashed horizontally
hardy yew
Ahh its only one song. Im making dataset. I remember there was talk about dc off...
hmm, i dunno actually if AI separation is prone to creating audio with DC offset
but if that's the case, then it's worth correcting for sure
slow aspen
but if that's the case, then it's worth correcting for sure
Yea thats what i remember people were saying. Is there a better way to send you a graph so you can see better? Also it was my first time trying RX 11 and there was dithering option in export section i leaved it on because I didnt want to screw something, should i disable that?
viral mason
what gpu do u have (Nvidia or AMD) and what are you planning on using it for?
I wouldn't immedietly go for the one that Erq sent bc if you have Nvidia u can use a much better one
slow aspen
I'm not sure what that is, I usually just truncate the silence in my datasets af...
Whats your way of cleaning dataset?
viral mason
Whats your way of cleaning dataset?
I mainly use the MVSEP site but I clean in the order of the ones shown in the video
too lazy to type all that
slow aspen
I mainly use the MVSEP site but I clean in the order of the ones shown in the vi...
oh thx ❤️
viral mason
no problem!
oh ok
for intel I can't help u
but are you sure it's not Nvidia?
are u using a laptop?
oh
yeah, for laptops it is
viral mason
I cannot help with intel, integrated gpus are kinda really bad for this stuff
you can use it still online but I have zero experience with the online alternative
Capy may be able to help tho
wrong user lol
the one typing is capy
slow aspen
hard to tell how it looks, as it's too squashed horizontally
made bigger picture. Does it look good?
hardy yew
Yea thats what i remember people were saying. Is there a better way to send you ...
TBH I don't think I'm competent enough to "rate" a dataset just by looking at spectrograms (unless there's something obvious, I guess). Another thing that it just needs to be listened to.
About dithering, it shouldn't affect the model much I think, but I've heard people smarter than me mention that RVC likes to learn the artificial noise introduced by dithering. I would imagine it's mostly for long-term training like pretrains... but dunno really.
Perhaps the better way is to stick to 32bit float
viral mason
someone like you (a normal not brain damaged individual) I can talk to normally
but that guy in there just didn't work correctly
hardy yew
made bigger picture. Does it look good?
that's still a bit too dense to read
I was playing a game, not typing for all that time xD
writing the message bit by bit
slow aspen
TBH I don't think I'm competent enough to "rate" a dataset just by looking at sp...
ok thx. Oh i cant really find 32 bit version of songs for my dataset, only 16 bit so i dont really have choice
viral mason
almost one everyday, kinda varies
hardy yew
ok thx. Oh i cant really find 32 bit version of songs for my dataset, only 16 bi...
but you can export from RX izotope into 32-bit float to avoid dithering at all
hardy yew
Capy may be able to help tho
And to address this, Intel's integrated GPU is not a thing that can be used with RVC really. The remaining options are:
a) inference on CPU, which is gonna be terribly slow
b) running RVC in cloud services
slow aspen
but you can export from RX izotope into 32-bit float to avoid dithering at all
Doesnt that add a fake data to a dataset that make model worse?
viral mason
that's why we have Nick, he pretty much auto deletes them on site once they admit to it
such a good mod
hardy yew
Doesnt that add a fake data to a dataset that make model worse?
if RX suggests dithering the output, that already means that you probably applied some processing that caused it to convert the audio to float in the first place. Anyway, if you do it on the entire dataset, it should be fine
Either way, dithering should also be OK
Just pointing out that more experienced people mentioned that it can be an issue, occassionally
slow aspen
Okk thanks very much
hardy yew
but not a huge one, i think
slow aspen
if RX suggests dithering the output, that already means that you probably applie...
Yea after resampling, dither option is showing up
viral mason
cyaa
slow aspen
👋
hardy yew
Yea after resampling, dither option is showing up
resampling is another thing, what causes RX to suggest dithering is when the loaded audio is in float and you're trying to export to a quantized format
that's why if you want to keep the data untouched you might want to just export 32bit float to keep the precision
slow aspen
ohh interesting
slow aspen
After de reverb?
viral mason
usually before
slow aspen
okk
viral mason
right after cleaning from the music
slow aspen
Thank you
viral mason
you're very welcome ^^
slow aspen
Last time i trained a model was in 2023 so im a little rusty xd
viral mason
oof
main things u don't need to do anymore is slicing audio to 3 seconds, I'd highly recommend using the new legacy core 1.6 pretrain and also use Applio
slow aspen
main things u don't need to do anymore is slicing audio to 3 seconds, I'd highly...
uuuu nice
hushed sinew
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc o...
pc gpu rx 6800XT 32 gb
os windows 11
aicovers
ebon tapir
For Vonovox just use block size at the default 0.30 or go up to 0.50 if you want
Ok but do virtual cable wtv matterss if they are diffrent bc when i use this it glitches sometimes
like it sounds like straight up ai glitchey
viral mason
Ok but do virtual cable wtv matterss if they are diffrent bc when i use this it ...
if you have no virtual audio cable it will not work at all in games or discord
ebon tapir
if you have no virtual audio cable it will not work at all in games or discord
No like i have a virtual audio cable but like do different ones affect the sound?
Cuz maybe i have a bad one thats why it sucks.
viral mason
No like i have a virtual audio cable but like do different ones affect the sound...
no they do the same thing, vac lite is better than VB cable tho as vb causes weird issues sometimes
ebon tapir
no they do the same thing, vac lite is better than VB cable tho as vb causes wei...
Do u know any good voice models mine sounds like a robot lowk
my soldier boy plan didnt work
😞
viral mason
is there any way u can show me what is happening?
ebon tapir
is there any way u can show me what is happening?
Its like
When i say small words
Its like
Mahsia
Sometimes
And it js doesnt some that much real
Sound
viral mason
I don't get what issue you're having :(
ebon tapir
okay so basically
Its ai
It doesnt sound whatever like the thing that they posted
Like they post sounds when u go on a model
Like theirs smooth
Mines ahh
viral mason
are you able to send a screenshot of your voice changer?
maybe a video of how it sounds?
ebon tapir
oh well idk when i record the video my sound usually dont go
ebon tapir
maybe a video of how it sounds?
Is there like other setting in vonovox to make it better or js pitch
nocturne mural
viral mason
nah, u can just make em urself for free
viral mason
Is there like other setting in vonovox to make it better or js pitch
pitch, and block size
that's it
do u have a loud fan or something in the background
anything your mic could be picking up?
ebon tapir
it js sounds ai like i dont know how to explain
i turn of the fan
cuz when i make it hot its hot and make it cold its cold so
but i think maybe its js the voice models but when i look at the voice model like there sounded good
When they post the sound of it
it sounds realistic asf
do u know any good voice models
viral mason
do u know any good voice models
I make good models and so does the guy who made the soldier boy model
viral mason
idk what you're looking for
ebon tapir
anything realistic
i can be a kitten js something realistic 
Icl soldier boy tuff idc if that sometimes glitches
ebon tapir
Soo
low shard
pc gpu rx 6800XT 32 gb
os windows 11
aicovers
MMVC is related to original wokada, which is for realtime voice changing, not ai covers, so you confused the programs up
hallow thistle
i can be a kitten js something realistic <:anime_pray:1159685390156967936>
Check out my Neco-Arc model in #1175430844685484042 
proven hill
has the hall of shame been removed once again?
it was filled with 90% of new users
no wonder why they removed it
viral mason
it was filled with 90% of new users
I used to to track weirdos
shaming them for misuse of rvc
proven hill
shaming them for misuse of rvc
who doesnt miseuse rvc
proven hill
I mean those "egirl trollers" <:joe_weird:1159569750377119774>
hello i need 3 latina 5 french 2 mommy
viral mason
hello i need 3 latina 5 french 2 mommy
viral mason
i have a rx 7800 xt
hallow thistle
40k, 32k only models
Use Discord search bar.
amber venture
if i'm starting fresh, what certs do I get for AI?
hallow thistle
if i'm starting fresh, what certs do I get for AI?
Information Technology. 
ebon tapir
I got u dw https://github.com/tg-develop/voice-changer/releases/download/b2397/v...
do u know what settings i use for soldiewr
deft depot
hi i need help
I’m having some performance issues with gemma-4-E4B-it-Q4_K_M.gguf
I’m running an RTX 3060 Ti, but the performance is still super slow. Here are some details:
The model is stored on SSD
Nvidia’s overlay (Alt+R) shows 60% GPU usage, but Task Manager only shows 10% (i think iknow why but not sure)
CPU usage is basically zero, and it’s only using about 3GB of RAM.
In the system info, it says: gemma4-manual:latest | 7.7 GB | 100% GPU | Context: 4096
The file size is 4.95GB, but it shows up as 7.7GB in the process
Is it bottleneck or what im confuesed? or is there something wrong with my GPU settings or dependencies pla pla pla idk? anyhelp would be appreciated
its q4bit
ollama_version=0.21.0
oh wait is this server about voice models omg
wrong server lol
proven pecan
So I ran dereverb + de-noise in UVR5, which worked pretty great.
trained the model in Applio
But two issues in Vonovox.
- Even with my normal voice, it still is laggy and slightly clicky on my RTX 4070 TI Super no matter the settings I change.
- it's completely unusable with my dad's damaged voice which sounds like a loud whisper.
My dad had throat cancer and one of his vocal cords was removed about 10 years ago.
I have about 45 minutes of his pre-damaged voice, it's pretty good, but not a modern high quality recording like you'd hear for audiobooks or anything like that. Still, not bad for the early 2000's.
I'm trying to figure out solutions for when he speaks publically.
But so far, AI doesn't seem to offer any 🙁
Does anyone have any ideas of what I could try? Anything to beef up the voice for live speech?
viral mason
do u know what settings i use for soldiewr
There's no settings you could change in Vonovox I'm not sure what's wrong with your voice changer 💔
I would suggest at this point to join the official discord server for Vonovox and ask for help there
viral mason
So I ran dereverb + de-noise in UVR5, which worked pretty great.
trained the mo...
You don't need to use denoise anymore unless the mic noise is bad, tbh I'm unsure what you could do if his voice irl is more of a whisper it most likely wouldn't work well on any realtime voice changer
What does this mean
What's your PC gpu (Nvidia or AMD) and did you get your voice changer from a YouTube tutorial?
Your voice changer is outdated then, I'll get you the one you need
Btw what are you using it for just curious
Are you using the voice of valorant characters to troll?
Ok
viral mason
I got u dw https://github.com/tg-develop/voice-changer/releases/download/b2397/v...
You'll need these two downloads
I'd recommend deleting the old voice changer you have now and using the new one you just downloaded so there's nothing conflicting
Btw for the second link (vac lite) run setup64 then install driver
And for the voice changer run mmvcserversio
viral mason
Uhm
simple ore
did not select a model
viral mason
oh
did you import a voice model?
it will not run if there is not a voice model in the program
all of those voices work
are you able to send a screenshot
a moderator would have to give you that ability
or you would have to level up by talking here
why egirl?
egirl models are usually used for catfishing/scamming
I wouldn't use them because of that reason
spongebob or stuff like that is good
what do you have your audio settings at?
my input is my headset microphone, and my output is line 1 (vac lite)
and you have this set to your AMD gpu?
ah that may be why
that part doesn't matter lol
just make sure your gpu is there and not CPU
under processing unit
uhm
do you have the model you imported selected like this?
this says intel, not AMD
are you sure you have an AMD gpu?
oh damn
intel really can't do this stuff locally very well at all I'll be honest
slim ivy
hey guys i need help
First of all, the Kaggle version of Applio needed to save a PTH file every 50 epochs, but for some reason it isn’t saving the PTH files. What’s the problem?
viral mason
First of all, the Kaggle version of Applio needed to save a PTH file every 50 ep...
follow this video I made
you may have somehow messed up a step
btw saving every 10 or 5 is much better than 50, that is excessive
slim ivy
you may have somehow messed up a step
i've checked the document from https://docs.aihub.gg/
my phone number is verified
i only have index file but not the pth.
viral mason
are you able to screenshot your training settings?
did you enable anything that was not selected by default?
slim ivy
are you able to screenshot your training settings?
yea but i cant sent the screenshots here
viral mason
ah
viral mason
sure yea
bold siren
Y'a des français ?
frail yoke
what should i use instead of weights or replay?
is there a google colab related page that can help me?
what can i use to train rvc models instead of weights since it shut down?
viscid moss
Is Applio still running on Colab? Im facing Colab disconnections in my notebooks... smh
whole carbon
Jak zrobić model AI?
viscid moss
pls keep it english only
fiery juniper
is it better to train rvc v1 or v2 in applio for w okada
viral mason
is it better to train rvc v1 or v2 in applio for w okada
V2, v1 is old and I don't think it's possible anymore to make v1
This is very old, what is your pc gpu (Nvidia or AMD) and what are you using it for, just curious
analog path
Hey, audio engineer here specializing in RVC training and dataset cleaning. Happy to help anyone with voice model questions
fiery juniper
V2, v1 is old and I don't think it's possible anymore to make v1
my gpu kinda ass, is 30 epoch enough..?
i tried training until 500 but it takes 10 minutes to do 1 epoch
is there a way i can continue off 30 epoch or... when i abort it
i have to start over (i use applio)
sorry im kinda new to this gsdhgsd
next knoll
do i put added in model or index, same thing with .pth?
viral mason
my gpu kinda ass, is 30 epoch enough..?
i tried training until 500 but it takes ...
You could just train online using Kaggle instead, it's much easier than using it locally with a bad pc, way over 100 epochs will always be needed for any model
viral mason
do i put added in model or index, same thing with .pth?
No need to edit the name of the file
exotic ridge
hello! I can't seem to inference nor tts in applio, it keeps showing error
wanton saffron
where did weight's models gter achived at/
low shard
where did weight's models gter achived at/
No RVC Model from Weights.gg/.com was archived, it would have taken too much storage and time
But weights took models from #1175430844685484042
low shard
who doesnt miseuse rvc
AI misuse is severly banned here, which includes those "e girl trollers catfishers".
proven hill
AI misuse is severly banned here, which includes those "e girl trollers catfishe...
i rarely see someone use it normally
low shard
how do i fix delay
This is a General AI Discord Server, please elaborate:
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
low shard
oh wait is this server about voice models omg
no this server isn't only about voice models, it used to be RVC-only, but now it's expanded to more things, it's just that unfortunately there's still a bit of RVC focus because of many people joining from old youtube tutorials
low shard
I’m having some performance issues with gemma-4-E4B-it-Q4_K_M.gguf
I’m running a...
- close other heavy apps
- what is the context window? maybe try to short it up
- have you tried checking with nvidia-smi in a CMD?
low shard
wrong server lol
you can ask everything AI related in this server dw
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
low shard
what should i use instead of weights or replay?
what's your pc gpu and os?
low shard
Is Applio still running on Colab? Im facing Colab disconnections in my notebooks...
like mid running? is it still happening? with the UI Google Colab?
@nocturne mural you might want to see this
low shard
my gpu kinda ass, is 30 epoch enough..?
i tried training until 500 but it takes ...
epochs are just a measure of how many times the model trained on the whole dataset, they aren't related to quality
what's your pc gpu and os?
low shard
i rarely see someone use it normally
everyone who uses it for misuse and bad things like NSFW get banned from here and don't get help, unfortunately those are related to the old ass youtube tuts that we hate and banned too
i wish those youtube tutorials never existed smh
proven hill
i wish those youtube tutorials never existed smh
same
fierce bone
All i see is that once I'm trying to convert a 3 second test vocal with any mode...
Just wondering if someone could shed another light on this. After getting Applio 3.6.2 to run in Windows 11, and confirmed that it sees the AMD card "RX 9070XT" under "Advanced Settings" in the" Training" tab, I went on downloading a few models just to test infering. "ie, Spongebob, TF2 Heavy in this case". I wanted to convert a 3 second sample saying "This is a sample audio for you. Do you like this model?", even tried other audio files. I do see in Task Manager under the "Performance" tab, the VRAM raising up when starting the conversion. But when I see something like this, I can't help but wondering if it actually is doing anything.
CPU: 5600x / GPU: RX 9070XT
- I'm following this guide: https://docs.aihub.gg/rvc/local/applio/ / Section: AMD on Windows (Precompiled Fix)
simple ore
Just wondering if someone could shed another light on this. After getting Applio...
what does it show in the console window?
with 9070xt you dont need to follow the zluda guide
fierce bone
with 9070xt you dont need to follow the zluda guide
Thing is, if i dont, it wont show up
fierce bone
what does it show in the console window?
Well, the screenshot was made yesterday and is closed. But I started another one, and this has just been sitting like that for roughly 600 seconds so far.
simple ore
Thing is, if i dont, it wont show up
with 9070xt you uninstall torch libraries, then you do env\python -m pip install torch torchaudio torchvision --index-url https://rocm.nightlies.amd.com/v2/gfx120X-all/
and after that is done you use normal run-applio.bat
frail yoke
No RVC Model from Weights.gg/.com was archived, it would have taken too much sto...
I looked for a Storm King model, previously from Weights, in here
fierce bone
and after that is done you use normal run-applio.bat
Well, that did something. I'm getting the following output this time.
safe harbor
Guys, can someone share a Mega/Google Drive link for Big Baby Tape RVC model? Weights link is dead
eager quest
I tried running the W-Okada Voice Changer through the official Google Colab linked in their official Github on my Windows PC, but shortly after running "Clone repository and install dependencies" inside the Google Colab while having the GPU selected as the runtime, it fails with this code output (this is not the whole code but I can't make the message too long. The part of the code you're seeing is the one after those "/sbin/ldconfig.real:" outputs):
(
Installing pre-dependencies...
ERROR: Could not find a version that satisfies the requirement faiss-gpu (from versions: none)
ERROR: No matching distribution found for faiss-gpu
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 261.0/261.0 kB 6.8 MB/s eta 0:00:00
Preparing metadata (pyproject.toml) ... done
Building wheel for pyworld (pyproject.toml) ... done
Installing dependencies from requirements.txt...
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 10.7/10.7 MB 97.8 MB/s eta 0:00:00
Installing build dependencies ... done
error: subprocess-exited-with-error
× Getting requirements to build wheel did not run successfully.
│ exit code: 1
╰─> See above for output.
note: This error originates from a subprocess, and is likely not a problem with pip.
Getting requirements to build wheel ... error
error: subprocess-exited-with-error
× Getting requirements to build wheel did not run successfully.
│ exit code: 1
╰─> See above for output.
note: This error originates from a subprocess, and is likely not a problem with pip.
Successfully installed all packages!
)
Can anybody help please?
hallow thistle
I tried running the W-Okada Voice Changer through the official Google Colab link...
What is your PC GPU? And what do you use the voice changer for? Because it looks like you're trying to install the program by yourself, but then the original W-Okada repository has been outdated.
queen token
why is my thing echoing
eager quest
What is your PC GPU? And what do you use the voice changer for? Because it looks...
Hello there! My PC GPU is a GTX 1650- not the best, but it meets the minimal requirement as I've already informed myself (I think). I would simply use the Voice Changer to troll some of my friends and generally just have fun with it like most do. I have no idea what you meant by "running the program myself" sadly as I'm not known with anything like that- but I'm actually following a tutorial from YouTube, and it worked for them- but somehow not for me. I followed every step exactly and it still doesn't seem to work.
hallow thistle
why is my thing echoing
What is your PC GPU? Did you follow any tutorial or guide before? And what do you use the voice changer for?
hallow thistle
Hello there! My PC GPU is a GTX 1650- not the best, but it meets the minimal req...
Why trolling?
eager quest
Why trolling?
Why does it matter exactly? And I don't really know if I have to answer that. And that doesn't help my problem at all. By trolling I meant having fun with my friends, since they understand my humor
queen token
What is your PC GPU? Did you follow any tutorial or guide before? And what do yo...
its 5090
why?
and why does it sound so ahh 😭
hallow thistle
Good question. My question, especially "what would you use the voice changer for", matters, in case to avoid providing help to use the program in bad ways. What about GPU? This also matters because this unit in your PC is used to process the voice changer audio. 
queen token
optimus prime roleplay
hallow thistle
Why does it matter exactly? And I don't really know if I have to answer that. An...
Now what about you?
eager quest
Now what about you?
I am sorry if I misunderstood. Now to my GPU, I have already said my GPU. My GPU is a NVIDIA GTA 1650- not the best as I said, but it meets the minimum requirement
hallow thistle
I am sorry if I misunderstood. Now to my GPU, I have already said my GPU. My GPU...
This has already answered. The another question, what do you use the voice changer for? E-girl trolling or what?
hallow thistle
optimus prime roleplay
Vonovox and Tg Develop's W-Okada are only known voice changers that can work with GeForce RTX 50 series; anything older than these might not work on RTX 50.
eager quest
This has already answered. The another question, what do you use the voice chang...
What are you trying to achive with this? I have already said I wanted to have fun using it and testing it, so please do not try to accuse me of any weird actions like these. No further comment on that. I am trying to respectfully and peacefully fix this problem by getting help.
hallow thistle
What are you trying to achive with this? I have already said I wanted to have fu...
Don't take my words too deep. Anyways, as what I said earlier, the original W-Okada is outdated. Use Vonovox or Tg Develop's W-Okada fork.
fiery juniper
epochs are just a measure of how many times the model trained on the whole datas...
uhh i use windows 10 and my gpu is a gtx 1060 6gb
proper hound
Does the rx7900XTX work?
queen token
Vonovox and Tg Develop's W-Okada are only known voice changers that can work wit...
i have vccd
its still echoing
hallow thistle
uhh i use windows 10 and my gpu is a gtx 1060 6gb
While this GPU is one of minimum possible for simple inferencing, to train a voice model faster I'd go for online websites like Kaggle and Google Colab.
hallow thistle
i have vccd
Which?
hallow thistle
hallow thistle
https://huggingface.co/wok000/vcclient000/tree/main
Well, that's the original W-Okada. Not recommended to use because it's old.
queen token
may you send me the other one
hallow thistle
-realtime
patent trellis
hallow thistle
queen token
do i download all of that stuff
hallow thistle
Guide docs exists, by the way. Download https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.001 and https://github.com/tg-develop/voice-changer/releases/download/b2397/voice-changer-windows-amd64-cuda.zip.002 to your download folder, use WinRAR or 7-Zip to open the .zip.001 one.
hallow thistle
Does the rx7900XTX work?
Works with what?
proper hound
The voice Changer
hallow thistle
The voice Changer
Yes, W-Okada "DirectML" does work with AMD Radeon GPU, but there's only specific version that can work. What do you use the voice changer for?
hallow thistle
vcclient_win_dml_2.1.4-alpha
That's the beta version of the original W-Okada. Not really recommended to use.
proper hound
What version do i need?
hallow thistle
Tg Develop's voice changer fork.
queen token
Guide docs exists, by the way. Download https://github.com/tg-develop/voice-chan...
it errors when i try uncompress
hallow thistle
it errors when i try uncompress
You sure you're using WinRAR or 7-Zip?
hallow thistle
low shard
I tried running the W-Okada Voice Changer through the official Google Colab link...
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
hallow thistle
voice-changer-windows-amd64-cuda.zip.002
Open the .zip.001, not .zip.002.
low shard
uhh i use windows 10 and my gpu is a gtx 1060 6gb
You just want to train an RVC Voice Model right? Try searching for the Applio Kaggle Cloud in the AI Hub Docs, your GPU might be weak
low shard
Does the rx7900XTX work?
For what? This isn't a voice model server, it's a general ai discord server
low shard
its 5090
This is a General AI Discord Server, please elaborate:
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
proper hound
Experiments with AI
fierce bone
Experiments with AI
.. AI is not a specific thing. You gotta have something you want to go for. 😛
proper hound
Have Ollama and Try a couple thinks
low widget
does anyone know how people generate those deep fake videos u see, i have a 4070 on windows 11, is it possible to do it locally?
queen token
! C:\Users\user\Downloads\voice-changer-windows-amd64-cuda.zip: Unexpected end of archive
@hallow thistle
hallow thistle
"Artificial intelligence" isn't always about just retrieval-based voice conversion (RVC).
hallow thistle
! C:\Users\user\Downloads\voice-changer-windows-amd64-cuda.zip: Unexpected end o...
If you have "Hide extensions for known file types" enabled in your Windows Explorer, you should disable it. Otherwise, your Explorer tricks you that there's "voice-changer-windows-amd64-cuda.zip" when it's actually "voice-changer-windows-amd64-cuda.zip.001", confusing things up.
fierce bone
Have Ollama and Try a couple thinks
Ollama is a tool for running language models locally. There are many use cases for that. Which one would be yours?
proper hound
I dont have a set goal i only want to Test and try a couple thinks there is nothing more behind that i got the thing i was here vor working THX
hallow thistle
I'm not really into vibe coding, but when it's about large-number maths it's gonna be fun. 
queen token
If you have "Hide extensions for known file types" enabled in your Windows Explo...
same issue
nvm
works
thank youuu
simple ore
Well, that did something. I'm getting the following output this time.
lemme ask someone, this should not happen
@fierce bone could you make a new environment variable, give it a name MIOPEN_FIND_MODE and value FAST
then restart applio
fierce bone
then restart applio
Done - restarted and still got similar output.
tried adding env variable both in user and system. Not simultaneously.
fierce bone
just one place
What would you suggest that I do from this point?
simple ore
i waiting for a response
low shard
Have Ollama and Try a couple thinks
Ollama supports rx7900XTX on both Windows and Linux iirc
low shard
optimus prime roleplay
are you sure? you seem to have a fresh new discord account with a girly pfp
are you perhaps doing e girl trolling / catfishing?
fiery juniper
You just want to train an RVC Voice Model right? Try searching for the Applio Ka...
uhhh, CAN WE MOVE TO DMS FOR THISS, I HAVEE ALOT OF QUESTIONS gdshgds
if its okay
hallow thistle
uhhh, CAN WE MOVE TO DMS FOR THISS, I HAVEE ALOT OF QUESTIONS gdshgds
if its oka...
Does it have to be private though? Generally, how to train a model sounds more like a general knowledge everyone to know.
nocturne mural
like mid running? is it still happening? with the UI Google Colab?
<@98977238850...
Well, it’s well known that Google Colab has ways to track your usage, and maybe they flag you for something? I’m not sure. At most, they reduced my session time from 3 hours to just an hour and a half, and eventually, I just got restricted from using Colab entirely, though I can still use other services.
fiery juniper
Does it have to be private though? Generally, how to train a model sounds more l...
well-- i just think ill flood the channel with like
a ton of question
BUT WE CAN DO IT HERE
im just wondering how the kaggle thing works 😭
and what epoch do i need for a 28 minute audio file
whats empty file thing
do i need it
whats refinegan
how do i make sure kaggle doesnt turn off while its training
hallow thistle
Wow, that's horrifying, but better make things simple.
fiery juniper
uhehgag yeah 😭 im sorry
idk how kaggle workss, wdym download, i thought its cloud 😔
hallow thistle
This is the main page for Kaggle.
fiery juniper
where do i uhh get to the training model thingy
sorry for bothering you rn mann T-T
hallow thistle
For a 28-minute audio dataset, usually I go for around 250- 450 epochs because RVC voice models often sound good on that epoch range at least to me; anything beyond 600 or 1000 epochs is too overkill and overtrained.
nocturne mural
how do i make sure kaggle doesnt turn off while its training
Well, it’s basically just interacting with the page every so often. I don’t have an exact timeframe for it, but I was doing it every 30 minutes and the session would last up to 6 hours. I also ran a script to automate the interaction; even though it would fail in the console, it somehow kept the session active anyway.
fiery juniper
Well, it’s basically just interacting with the page every so often. I don’t have...
WOAHHH
i love you guys, please tell me more GDSHSAHDSA
IM IN THE PAGE NOWW
hallow thistle
To keep Kaggle (or Google Colab) page tab running when I switch to another tab, I often do this trick.
Do I need to answer all of these questions?
nocturne mural
To keep Kaggle (or Google Colab) page tab running when I switch to another tab, ...
https://tenor.com/view/cat-eyes-eyeballs-cat-eye-emoji-eye-emoji-cat-looking-gif-9406810929029867845
fiery juniper
Do I need to answer all of these questions?
im sorry gdasdhsa
uhh where is like
the page where i can train..
nocturne mural
the page where i can train..
fiery juniper
hallow thistle
Don't be too hurry. On Kaggle, go to Code, click "Import Notebook", copy the link "https://github.com/IAHispano/Applio/blob/main/assets/Applio_Kaggle.ipynb" and paste into "Enter URL" input, and then click Import button.
fiery juniper
oh its doing something
hallow thistle
The guide docs on how to use Applio RVC on Kaggle exists, by the way. https://docs.aihub.gg/rvc/cloud/applio-cloud/#kaggle
fiery juniper
The guide docs on how to use Applio RVC on Kaggle exists, by the way. https://do...
oooh
hallow thistle
On your right side, there's "Session options" section. Set these like mine.
fiery juniper
is it just too fast or-- the thing ran into sum error-
viral mason
https://cdn.discordapp.com/attachments/1159290139609137264/1495102876370997319/2...
My video btw ❤️
nocturne mural
My video btw ❤️
viral mason
is it just too fast or-- the thing ran into sum error-
You should watch the video pan sent, it's very helpful
You'll see it if you scroll up a bit, I made it very simple and easy to understand how to use Kaggle's applio
nocturne mural
is it just too fast or-- the thing ran into sum error-
Maybe it's an issue with the Python version being used in that notebook; I completely forgot that Kaggle is still using 3.11 when we're already on 3.12.
@viral mason Is Kaggle working for you? I've never seen a Keras error before.
nocturne mural
is it just too fast or-- the thing ran into sum error-
.
viral mason
Not sure, haven't been awake long enough to test it today
I was using fine last night tho
low shard
Well, it’s well known that Google Colab has ways to track your usage, and maybe ...
fair, just added more warnings about it
low shard
uhhh, CAN WE MOVE TO DMS FOR THISS, I HAVEE ALOT OF QUESTIONS gdshgds
if its oka...
i forgot to check this, but its way better to ask there where multiple people can answer you
nocturne mural
Where do I put that
Well, I guess it's at the top of the cell where Applio runs. Though if you say you aren't having any issues, then it might only happen with newer versions of TensorBoard, and your environment probably has an older one that isn't forcing a 'new' Keras version for now.
viral mason
Well, I guess it's at the top of the cell where Applio runs. Though if you say y...
is this only an error for tensorboard or does it also effect applio?
nocturne mural
is this only an error for tensorboard or does it also effect applio?
Well, only to TensorBoard, since Applio on the main branch doesn't have this issue.
azure ridge
would anybody be willing to help me on a problem? I have found this video which I suspect has AI generated audio but I simply cannot tell. If anybody thinks they may be able to tell could they help many thanks 🙏
left dirge
Detailed Description of the Problem:
I have a weird error where my client audio simply doesn't work, there's no yellow warning error, I can use it and select my line and microphone, but when trying to listen to the audio, there's no output, it just doesn't produce sound, no matter how I try to use client it simply doesn't produce sound.
Using the latest version and AMD. I've tried reinstalling, using different versions, using a different browser, etc. I can't get the client audio to work, this happened around a year ago- But since I couldn't figure out a way to fix it I came here.
So far I had been using the server option whenever I wanted to mess with the app, but the lack of noise supression is a pain I can't ignore.
Full GPU Name: AMD Radeon RX 6800
Operating System: Windows 10
Screenshot:
viral mason
`Detailed Description of the Problem`:
I have a weird error where my client audi...
yeah I've got that same problem where client just doesn't work, no clue how to fix it
if you have an Nvidia gpu I'd suggest just switch to Vonovox
left dirge
AMD GPU as I said, so can't use any Nvdia option
viral mason
dang
left dirge
My main problem is the lack of noise supression, I tried using a separate noise supression app but I had no luck finding any that worked lmao.
viral mason
are you able to send a recording of what it sounds like with no noise suppression?
it can't be that bad right?
left dirge
are you able to send a recording of what it sounds like with no noise suppressio...
I could- If I had any app to send a recording. But it's horrible, there's bumps, my breathing, it feels like someone is knocking on my microphone even if I'm not speaking, I've got noisy neighboors too so it's just impossible to use at all.
viral mason
ah
yeah having literally anyone else but you go through the mic makes it impossible to use
if there's multiple people it just becomes a mess
left dirge
Yep. I don't really have a noise free enviroment I can use so client usually worked perfectly for me but since it stopped working it's been a pain.
viral mason
I use a complicated setup that has voicemod, and fl studio, plus a bunch of virtrual cable stuff with my voice changer to add noise suppression but idk if that would work with your stuff
here's the download for Vonovox, and vac lite is for connecting the voice changer to discord or any games u play
if u need any help just lmk
formal ermine
ohh thank you so muchhh
viral mason
you're welcome!
hollow valley
the link
low shard
is there a specific reason why? there are multiple people that can help you there
low shard
why isnt this workingß?
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
there are a million reasons why something might not work
fallen temple
I’m trying to clone a voice. I already have the .wav file, but I’m missing the pre-trained model. Which one is the most powerful as of today? Which one clones the voice most accurately?
viral mason
sure
gusty socket
Hey, I'm trying to setup tg-okada on linux and amd gpu, it sorta works, but uses cpu isntead of gpu and there's no option to switch. How do I fix this?
viral mason
Hey, I'm trying to setup tg-okada on linux and amd gpu, it sorta works, but uses...
you might of downloaded the wrong version, I don't know which one is for specifically linux tho sadly
ping the helper role
formant is only a pitch shifter, no need to change that normally
gusty socket
you might of downloaded the wrong version, I don't know which one is for specifi...
oh, got it. what fork would you recommend? I'm reading wiki, there are also applio and vonovox, as well as 3 okada forks
proven hill
what do you need it for?
viral mason
don't use any of those :(
soldier boy or literally anything but those kinda voices are fine
people don't use em for that they use them to steal money and get free stuff ect
it's not technically against the rules here to make those voices sadly
proven hill
and they got banned
viral mason
damn
viral mason
i need help
with whaaa?
viscid moss
like mid running? is it still happening? with the UI Google Colab?
<@98977238850...
yeah, installation works. But after a min... They kick u out, there's no words like "RVC", "WebUI" into the notebook. So...... I though they are checking the file names and code inside those files bcuz I spent some credits on Gemini to refactor the variables, imports, files, etc and it worked.
proven hill
I am already taken dear friend I'm sorry
i didnt mean to ask you out 😭😭😭😭😭
youre too young
i wanted help in finding love in general lmfao
viral mason
i wanted help in finding love in general lmfao
oh lol
I'm not that young but idk how old u are so maybe I am for u
proven hill
I'm not that young but idk how old u are so maybe I am for u<:misc_shrug:1420419...
im very old
viral mason
you'll find someone Ilaria
hardy yew
get a doggo, that's true love
viral mason
pets are true friends
gusty socket
Has anyone on linux and amd gpu installed any of okada rvc successfully? My only braincell is fighting a losing battle right now
fallen temple
Help, everyone: I'm trying to clone a voice. I already have the .wav file, but I'm missing the pre-trained model. Which one is currently the most powerful? Which one clones the voice most accurately?
hardy yew
Help, everyone: I'm trying to clone a voice. I already have the .wav file, but I...
OG is good, can also try something like Legacy Core 1.6
fallen temple
OG is good, can also try something like Legacy Core 1.6
help
I'm using that one and it gives me an error:
Loaded pretrained (G) 'rvc\models\pretraineds\custom\Legacy_Core1.6_G_11.pth'
The parameters of the pretrain model such as the sample rate or architecture do not match the selected model.
Error(s) in loading state_dict for Synthesizer:
size mismatch for dec.ups.0.parametrizations.weight.original1: copying a param with shape torch.Size([512, 256, 20]) from checkpoint, the shape in current model is torch.Size([512, 256, 16]).
size mismatch for dec.noise_convs.0.weight: copying a param with shape torch.Size([256, 1, 64]) from checkpoint, the shape in current model is torch.Size([256, 1, 80]).
size mismatch for enc_q.pre.weight: copying a param with shape torch.Size([192, 513, 1]) from checkpoint, the shape in current model is torch.Size([192, 1025, 1]).
hardy yew
are you using the appropriate sample rate?
hollow lily
Hi guys! I'm looking for some assistance with an RVC conversion. I've prepared all the necessary assets: a clean vocal stem of 'Face', the instrumental, and a Scally Milano .pth model.
Since I'm having trouble setting up the inference environment, could someone please process the vocal for me? Important: I'm aiming for a very soft and gentle vocal style, similar to a love song. If you're running it, please try to keep the delivery smooth and emotive.
Alternatively, could you point me toward a stable, free WebUI where I can use my own model? Much appreciated!
gusty socket
Trying to install tg okada fork on CachyOS Linux with RX 6800XT GPU and getting this error:
ImportError: /MMVCServerSIO/_internal/onnxruntime/capi/onnxruntime_pybind11_state.so: cannot enable executable stack as shared object requires: Invalid argument
proven hill
get a doggo, that's true love
i have two dogs and a cat
viral mason
@low shard get him outta here
proven hill
<@911742715019001897> get him outta here
lmfao
@rotund hound i can help you with a real job
first we need an application
proven hill
what are you even using
left dirge
https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/#cli...
I tried to explain that it wasn't that error. It's not the yellow warning where it says "Client Audio not avaliable", I event sent a photo where you can see that doesn't appear. The issue has nothing to do with that known error.
next knoll
can anyone give good settings for the ai? i know theres like a document that has a bunch of settings on it for your certain specs
viral mason
can anyone give good settings for the ai? i know theres like a document that has...
settings for what exactly?
next knoll
settings for what exactly?
nvm i found it
hollow raptor
hey guys can someone help me and tell me why its not picking up my voice on discord
viral mason
hey guys can someone help me and tell me why its not picking up my voice on disc...
could u explain what you're trying to do? what's your pc gpu (Nvidia or AMD)
hollow raptor
is the voice changer only for Nvidia and and gpu?
viral mason
is the voice changer only for Nvidia and and gpu?
if you have Intel (usually a laptop) there isn't anything for it that can run well at all
hollow raptor
oh alr thx
hallow thistle
is the voice changer only for Nvidia and and gpu?
What?
hallow thistle
Hi guys! I'm looking for some assistance with an RVC conversion. I've prepared a...
There's Applio RVC.
tribal minnow
anyone wanna help build a stocktake system with claude can pay
floral void
204360
arctic sandal
Hey
simple ore
What would you suggest that I do from this point?
can you try with this rvc\lib\zluda.py
the compilation error is weird and I suspect it requires VC Build Tools installed
fierce bone
can you try with this rvc\lib\zluda.py
I'll check in a bit. Just got home.
fierce bone
can you try with this rvc\lib\zluda.py
That's good progress! Vocal and model sounds as expected after conversion now!
Now for training... I'm testing with one of my acapellas. This was at first try.
low shard
yeah, installation works. But after a min... They kick u out, there's no words l...
interesting, so refactoring every single file of your project bypassed it? on what did you try it? did you tell gemini to make it look like a weird obfuscated program?
low shard
<@911742715019001897> get him outta here
smh
viscid moss
interesting, so refactoring every single file of your project bypassed it? on wh...
Yes! Like this: https://github.com/Eddycrack864/AICGP12/commit/a0fc59a68184e225666ad32c9091ca536da0da20
renamed a lot of stuff and it worked : D
hardy yew
renamed a lot of stuff and it worked : D
viscid moss
I'll check the rest of my notebooks (UVR5 UI, CoverMaker)
crazy how they check what u are doing on Colab 
echo meadow
is there something like idk, i need something where is like text to speech
hallow thistle
is there something like idk, i need something where is like text to speech
Applio RVC has TTS feature built in.
simple ore
Now for training... I'm testing with one of my acapellas. This was at first try.
edit rvc\train\train.py and delete lines ``` dist.init_process_group(
backend="gloo" if sys.platform == "win32" or device.type != "cuda" else "nccl",
init_method="env://",
world_size=n_gpus if device.type == "cuda" else 1,
rank=rank if device.type == "cuda" else 0,
)
You are friggin crazy... This is looking very promising!
I can't thank you enough @simple ore ...
hardy yew
You are friggin crazy... This is looking very promising!
awesome! I've been lurking from time to time and reading about some of the struggles
looks like it paid off
fierce bone
Absolutely. I'm beyond impressed right now.
echo meadow
Applio RVC has TTS feature built in.
hellllllllllllllllllo
hallow thistle
may you can send the link please?
-rvc
echo meadow
nice helper
hallow thistle
If you use Windows, this is the direct download link for Applio RVC. https://huggingface.co/IAHispano/Applio/resolve/main/Compiled/Windows/ApplioV3.6.2.zip
simple ore
I can't thank you enough <@155030383648440320> ...
3.88s/it is super slow
even my old 6700XT was faster with Zluda
fierce bone
3.88s/it is super slow
Yea, but from the looks of it my cpu is either taking the load, or holding it back.
simple ore
99% CPU is not right
fierce bone
Anything there can be done about it?
simple ore
Anything there can be done about it?
install vc build tools, re-enable cudnn
echo meadow
If you use Windows, this is the direct download link for Applio RVC. https://hug...
i want collab
cant even download dione launcher
hallow thistle
i want collab
By the way. What is your PC GPU?
echo meadow
By the way. What is your PC GPU?
4070 super bro
proven hill
4070 super bro
more than enough for local use
hallow thistle
4070 super bro
You should definitely run Applio RVC locally. Dione Launcher is not needed.
echo meadow
You should definitely run Applio RVC locally. Dione Launcher is not needed.
it shows only blank when run applio runbat
viral mason
crazy how they check what u are doing on Colab <:gura_sideeye:142041974402816412...
This is why I love Kaggle
proven hill
it shows only blank when run applio runbat
what version did you download
viral mason
i want collab
Use it on Kaggle instead colab is bad, only 4 hours max compared to the 30 hours Kaggle gives
fierce bone
install vc build tools, re-enable cudnn
Re-enabling, as in restoring the train.py or how ?
nocturne mural
it shows only blank when run applio runbat
When Applio starts for the first time, it can take quite a while; however, it won't be as slow in subsequent runs
viral mason
What gpu do you have (Nvidia or AMD) and what are you trying to use the voice changer for?
What kind of trolling? Like playing as Goku or Darth Vader something like that?
That isn't allowed here <@&1159293140440723499>
A lot of people use girl voices to get free stuff in games or scam people
It's disgusting
Catfishing
fierce bone
zluda.py
Just by setting torch.backends.cudnn.enabled = True? Because I did that after installing vc build tools 2026 but the cpu still got worked up.
simple ore
Just by setting torch.backends.cudnn.enabled = True? Because I did that after in...
correct
viral mason
Weirdos or people that get tricked into believing the bad person using the voice changer is a female
simple ore
@fierce bone MSVC v14x, c++ cmake, and win10 or win11 sdk should be selected
viral mason
Both people would be in the wrong , tricking someone isn't right no matter what they're doing unless they like kids
Anyone who is like that should be sent to death
Just don't use it for trolling as a girl and you'll receive help
Maybe
Use Vonovox, the link is in this chat somewhere but I won't be sending it since I already have notified the mods about earlier
Just use the searchbar and look for it
Yup
Best one for Nvidia
Nah
There's nothing you'd need a tutorial about anyway
Just download and run the start file
viral mason
What is sonobus?
Whatever is in a yt tutorial is outdated
Just download what I said from earlier lol
Yea that
And the second link too
Vac lite
?
O
fierce bone
<@258608245894152193> MSVC v14x, c++ cmake, and win10 or win11 sdk should be sel...
Double-checked, should be installed. (Don't know why it shows two SDK's for Windows 11 in this case. After relaunching the installer, the 2nd instance was gone. "10.0.26100.7705" is the one installed rn.
void flume
my rvc suddenly doesnt work anymore
rvc stands for 'retreival-based voice conversion'
rustic pumice
rvc stands for 'retreival-based voice conversion'
bro
im talking about the voice changer obviousluy
viral mason
What gpu do u have (Nvidia or AMD) and what are u using it for?
void flume
No, it's not obvious since people use rvc for making models as well
rustic pumice
What gpu do u have (Nvidia or AMD) and what are u using it for?
bro i have a 5070 and im using it for fun
before it was working
as always
now after a sudden it doesnt work
viral mason
Get Vonovox
rustic pumice
Get Vonovox
ew that shi is ahh
rustic pumice
the sound quality is bad
viral mason
Guess I won't be helping you
rustic pumice
oki lazy boy
void flume
Not very nice
viral mason
Literally this is how Vonovox sounds in the most recent beta, it's fine
viral mason
oki lazy boy
Stinky
rustic pumice
took shower 2 times today and brushed my teeth 3 times today and shaved my dih
i have vonovox installed though but i forgot which one to run
fierce bone
Sounds very retrieval-based.
viral mason
i have vonovox installed though but i forgot which one to run
Do u have the most recent beta? If not u should download that instead of the current full release
rustic pumice
Do u have the most recent beta? If not u should download that instead of the cur...
im not sure
viral mason
Lemme get u the link in case not
rustic pumice
i had that installed since like months ago
then decided to not use it cause it was bad
rustic pumice
i have vac lite
viral mason
Peak
rustic pumice
https://huggingface.co/dr87/vonovox/resolve/c8034f5f6d50648a8109bb4f847182362e2b...
so if i get vonovox
ill lose my voice models
rustic pumice
idk how to do that
rustic pumice
idk how to transfer to vonovox
viral mason
Wdym
viral mason
Just put em in vonovox the same way you did with what u use now lol
rustic pumice
Just put em in vonovox the same way you did with what u use now lol
ik but the voice models folders are deleted besides the voice models inside the voice changer i used
void flume
w-okada does convert the originals into other files, now that I think about it
though, those files should be the same, despite having a different extension
viral mason
ik but the voice models folders are deleted besides the voice models inside the ...
You can drag the voice models (the pth and index) files out of there to save them
In the folders
rustic pumice
You can drag the voice models (the pth and index) files out of there to save the...
where can i find them
in the other voice changer
viral mason
Rumi help me out here
viral mason
Start.bat
void flume
Inside
MMVCServerSIO\model_dir
you'll find numbered folders
Each of them contains the model along with the profile settings stored by w-okada
rustic pumice
ty
rustic pumice
Each of them contains the model along with the profile settings stored by w-okad...
it has numbers tho idk which one i wanna take it
void flume
but, as I mentioned, rather than pth, they have been converted to .safetensors files
rustic pumice
can i js put them all
void flume
.safetensors and .pth are both safetensor files
rustic pumice
but, as I mentioned, rather than pth, they have been converted to .safetensors f...
so what can i do about that
how can i put them in vonovox?
void flume
so what can i do about that
You can possibly just import them into vonovox as is.
rustic pumice
You can possibly just import them into vonovox as is.
how do i turn it to pth
bro
it doesnt show anything
void flume
idk. I glanced at a model I imported to test to see what the difference is, and the two files are different.
the pth isn't the same as the safetensor one
w-okada is doing something to the model as such
@viral mason I know too little on AI and programming.
It looks like w-okada is using huggingface's library to convert PyTorch files into Huggingface's safetensor files.
there is a converter that they have for from pt to .safetensor, but no clue on the other way around. (can't find one)
Also I don't have an nvidia card, so I can't test whether or not vonovox supports safetensors natively
proven hill
will someone help a poor girl with a broken heart??
void flume
how do i turn it to pth
It's not that simple. pth is basically a container of pytorch data. You need to know how the model is assembled as pth, rather than just using any converter. idk if the original tool used to make it can do that, but- anyway, I think it's probably easier if you just redownload the model.
viral mason
<@1023278814752677918> I know too little on AI and programming.
It looks like w-...
Yikes
rustic pumice
It's not that simple. pth is basically a container of pytorch data. You need to ...
dude i made that model
and i dont ahve any saved things of it
only that way i can
void flume
Well, ... Vonovox's dev planned adding safetensor support in version 8 apparently, not this next update (version 7).
so if you want to use that model, you'd need to stick with w-okada for now
rustic pumice
so if you want to use that model, you'd need to stick with w-okada for now
w-okada doesnt even work dude
i used it many times
and it suddenly doesnt work anymore
doesnt even make sense bro wtf
void flume
There are a couple of reasons I can think of why that could be...
rustic pumice
There are a couple of reasons I can think of why that could be...
?
void flume
In your AppData/Local/Temp directory, w-okada tends to produce files that it uses for.. idk what actually
but it is audio related. You can find many copies of in.wav and out.wav
they're not discarded when the server is shut off because... they're in use at the time of shutdown.
so there is a permission issue there and blah blah- anyway.
It's possible its trying to grab one of the existing ones and fails because it no longer has permission to use that
You basically may want to try and find those and delete them before running the server. It may clean up space anyway
Windows Updates might have changed permissions on your device drivers for your browser, which causes issues with ... well audio arriving at the server.
(you may need to allow the browser to use the mic (again), permission wise I mean.)
if its not hearing anything it won't convert
If the entire server just crashes, then- idk, like- idk what your computer is doing basically.
You can always try to redownload w-okada
w-okada does produce logs. Perhaps reading them gives a hint on why it isn't working
vcclient.log
without information I can only go gamble as to why it stopped working and how you could maybe fix it
simple ore
@fierce bone here's something to try
set CC = "path-to/venv/Lib/site-packages/_rocm_sdk_core/lib/llvm/bin/clang.exe"
set HIP_PATH = "path-to/venv/Lib/site-packages/_rocm_sdk_core"
fierce bone
set CC = "path-to/venv/Lib/site-packages/_rocm_sdk_core/lib/llvm/bin/clang.exe"
...
Thanks. I'll give it a go tomorrow. 💪
viral mason
<@&1159293140440723499>
proven hill
<@&1159293140440723499>
pure eye bleach
paper bloom
hey everyone
i got the owakada ai voice changer idk whats it exactly called but im using the b 2332 eversion is that the newst one?
or are there better once now?
proven hill
what do you need it for
proven hill
i got the owakada ai voice changer idk whats it exactly called but im using the ...
what do you need it for
noble heath
after you have converted your audio on applio how're you supposed to download it?
burnt turret
Hi, I have a question: how can I publish my first voice model?
noble heath
I'm sorry but where is the download button??
burnt turret
How can I publish my voice model?
unique oar
closing
paper hinge
hello I use applio, how do I select voice model and index file ? I have them but I don't know how to put them inside
nocturne mural
hello I use applio, how do I select voice model and index file ? I have them but...
Applio/logs/mymodel/.pth . index
paper hinge
Applio/logs/mymodel/.pth . index
sorry I don't find mymodel folder
is this a mistake of me?
nocturne mural
sorry I don't find mymodel folder
'mymodel' is just a reference to your model's name.
hallow thistle
sorry I don't find mymodel folder
"mymodel" is an example folder name; the folder name can be anything (like Neco-arc).
paper hinge
"mymodel" is an example folder name; the folder name can be anything (like Neco-...
oh so I can put it straight like this?
hallow thistle
Extract your voice model files into /Applio/logs, refresh the Applio program, simply as that.
hallow thistle
How can I publish my voice model?
nocturne oyster
Hi, I'm new here. I just found out that Weights closed down and when I went to check the website, there's this replay to download. Does it work?
grizzled crystal
Hey Someone have the the link of the last version of Applio
hallow thistle
Hey Someone have the the link of the last version of Applio
Is your PC GPU still GeForce RTX 4060 or you got another one?
You could've answer in here instead of my direct message. This is the download link of Applio RVC. https://huggingface.co/IAHispano/Applio/resolve/main/Compiled/Windows/ApplioV3.6.2.zip
-rvc
meager jay
I’m an AI & Full Stack Engineer focused on building production ready AI systems, not just prototypes. Most of my work is around connecting LLMs with real infrastructure APIs, databases, tools, and business logic so AI can actually run reliably in real workflows. I usually work with things like: LLM systems & orchestration (DSPy, LangChain, AutoGen, CrewAI, ReAct), RAG pipelines with vector databases and custom retrieval Multi-agent systems with planning and tool use, Multimodal AI (Whisper, CLIP, YOLOv8, TTS), AI image / video generation pipelines Backend & full-stack (FastAPI, NestJS, Next.js, React), Automation & integrations (n8n, Zapier, Make, custom APIs)
hardy yew
hallow thistle
I’m an AI & Full Stack Engineer focused on building production ready AI systems,...
Don't try to say something like this. I need more of human response. Is there anything you'd like to get help with?
low shard
Yes! Like this: https://github.com/Eddycrack864/AICGP12/commit/a0fc59a68184e2256...
Send this in DMS so I remember plis
Elaborate more
low shard
That isn't allowed here <@&1159293140440723499>
Thx
Don't help at all people who ask that no matter what
low shard
im talking about the voice changer obviousluy
Its not obviously, RVC doesn't mean that, and this is a general ai discord server not a voice model one only
low shard
my rvc suddenly doesnt work anymore
Elaborate your PC GPU, os, if you're trying to do tts or ai covers or e girl trolling / Catifshing or roleplay, and the tutorial link
low shard
i got the owakada ai voice changer idk whats it exactly called but im using the ...
Elaborate more
Your PC GPU, os, what are you trying to do and the tutorial link