im here
#✨│ai-help
misty haven
misty haven
i did
misty haven
thanks
low shard
in model put the model.pth file
low shard
in index put the model index
brittle wing
any idea what could be wrong i can hear myself via monitoring but no audio comes out on my virtual cable
low shard
A VAC (Virtual Audio Cable) makes a fake audio device, used to re-route the audio of different programs
In Wokada context, it's used to get the output of wokada as the input in other programs
is line 1 the input in the other program ur using wokada with?
brittle wing
A VAC (Virtual Audio Cable) makes a fake audio device, used to re-route the audi...
yes
i use line 1 on discord or any other voice call program
and no sound is coming out of the mic
low shard
and no sound is coming out of the mic
did u do this
brittle wing
did u do this
yes, it was working before and just randomly stopped
low shard
yes, it was working before and just randomly stopped
did u uninstall vb audio cable
brittle wing
did u uninstall vb audio cable
no
brittle wing
Uninstall it
so just reinstall it ?
low shard
so just reinstall it ?
Vb audio cable and vac lite are 2 different programs
digital ember
keep getting "unhandledrejection" error on okada fork how can i fix this
low shard
vb audio cable is known for giving issues on windows, uninstall it and keep using line 1 from vac lite
low shard
keep getting "unhandledrejection" error on okada fork how can i fix this
elaborate
- what's your pc gpu
- what tut link are u using
- what browser are u using
the more info, the better
digital ember
rtx 3050 ti laptop
https://rentry.co/ForkVoiceChangerGuide
brave browser
it was working perfectly fine yesterday
low shard
rtx 3050 ti laptop
https://rentry.co/ForkVoiceChangerGuide
brave browser
it was...
try either using another browser (like chrome or firefox) or play around with the browser settings
digital ember
alright
digital ember
on start up
low shard
on start up
yeah it's just best to use chrome or firefox then
digital ember
yeah it works on firefox thanks for helping
low shard
yeah it works on firefox thanks for helping
yw
might be a brave updated that fucked it up
digital ember
might be
blazing panther
is this fork a new version
low shard
is this fork a new version
Wokada has 2 main versions:
- Original made by Wok
- Deiteris fork (modified version) made by Deiteris
each version has it's own updates
the latest deiteris fork has way better performance and quality than the latest original
blazing panther
if im correct, the last version of deiteris' was 5 months ago?
blazing panther
okay cool
grim perch
Hey where can i download the sample weights audios to use “this is a weights.gg ai voice model… etc”
waxen pike
Wokada has 2 main versions:
- Original made by Wok
- Deiteris fork (modified ver...
Why does a female model when I say something feel like she can't spell half the letters and has no teeth at all? Whispers.
low shard
Why does a female model when I say something feel like she can't spell half the ...
show a screenshot of ur wokada settings
waxen pike
low shard
set extra to 2.0 and try other models
waxen pike
set extra to 2.0 and try other models
all the women's models have this problem.
that's exactly what I've noticed with myself.
low shard
all the women's models have this problem.
does it like cutoff at the start or end of phrases
waxen pike
the model lisped like she had no teeth.
low shard
the model lisped like she had no teeth.
yes, try playing with the pitch and the suggested models
not all models are good
waxen pike
https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#models-to-t...
how do you train them to fit your voice?
low shard
how do you train them to fit your voice?
You can't
You train a model based on the voice of the person, not based on how to make it fit to work with your voice
That's why you need some voice acting too
low shard
You know a better model than the one you threw at me? psu2go is not perfect.
Nope, don't expect models to be perfect at everything like laughing
Laughing is an RVC limitations for example
You can just use #1175430844685484042 or https://weights.com to find other models
low shard
why the voice isn't working?
elaborate
- what's ur pc gpu
- what do u want to do
- what tut link did u use
- which game? and are the game graphics to the minimum
- what is the issue
- share a screenshot of the program settings
reply to all the above, the more info, the easier for you to get helped
marble thistle
can someone help with why this is playing back my voice but not with the voice changer, i have an amd pc so idk if its that, i used this tut link https://www.youtube.com/watch?v=pHhjg2JwdPI
low shard
can someone help with why this is playing back my voice but not with the voice c...
you got an old original version of wokada, which is prehistorical
uninstall all u got off youtube
what's ur pc gpu
marble thistle
you got an old original version of wokada, which is prehistorical
lmao ok
marble thistle
what's ur pc gpu
how do i check
low shard
how do i check
You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU
low shard
AMD Radeon RX6600
nice, uninstall ALL you got off video tutorials
only written guides are updated
-realtime
patent trellis
marble thistle
wydm fork
low shard
wydm fork
fork means modified version in IT
it's a modified version by Deiteris
read that guide and u will be all done basically
low shard
yw
marble thistle
is it supposed to open in the web
low shard
is it supposed to open in the web
brittle wing
-colab
stiff crystal
I'm trying to make an ai voice model but I can't get the colab to work.
olive plover
What is the difference between
FORK
RVC .?
stiff crystal
I'm having trouble generating an index for a voice model.
brittle wing
Hello, can someone here help me with problems With the Google coolab with the hina mod?
Good night
grim perch
What is models hub?
golden walrus
guys, is it possible to fine tune embedder ?
glacial pollen
guys, is it possible to fine tune embedder ?
Yes but it is not something a casual user would do or be capable of ( at least, easily )
@ dr87 might perhaps provide you some info or papers if you asked nicely
golden walrus
Yes but it is not something a casual user would do or be capable of ( at least, ...
thank you so much
glacial pollen
mh mh, sure thing 
golden walrus
should i dm him or tag him lah 😦
iron ermine
im having trouble with the ai voice changer
hallow thistle
im having trouble with the ai voice changer
!howtoask
hallow thistle
Which W-Okada are you trying to use? What is your PC GPU? And make sure you elaborate more about your issue.
hallow thistle
What is the difference between
FORK
RVC .?
What is this supposed to mean? Is it fork RVC, W-Okada or something else?
hallow thistle
how do i get the voice changer
What is your PC GPU?
full moss
nvidia
full moss
What is your PC GPU?
nvidia
idk about pc so i may be wrong
but i use nvidia
the graphics
hallow thistle
To check your PC GPU, open Task Manager, go to Performance tab, spot where GPU 0 or GPU 1 is in the left side, and click one of it to reveal its full name on right side.
hallow thistle
Click on GPU 0.
hallow thistle
Task Manager will say something like NVIDIA GeForce RTX 4090 in big text.
hallow thistle
nvidia geforce rtx 4070 laptop gpu
Download and use this W-Okada from this guide. https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows
full moss
ok
where do i download
in windows? that text?
Virtual Audio Cable
#A Virtual Audio Cable (VAC) is what you need to use the voice changer on Discord & Games.
that?
hallow thistle
in windows? that text?
full moss
ok
hallow thistle
Virtual Audio Cable
#A Virtual Audio Cable (VAC) is what you need to use the voi...
That's Virtual Audio Cable, where you need this program to use between W-Okada and Discord or any voice call program.
hallow thistle
!howtoask
hallow thistle
That's true. But I did !howtoask in case if he asks where to download link or little step any further.
strange mountain
my speaker cant work when i use voicemeeter for Deiteris' W Okada Fork? (i find many source but cant fix )
hallow thistle
my speaker cant work when i use voicemeeter for Deiteris' W Okada Fork? (i find ...
Use Virtual Audio Cable lite instead. VB-Cable and Voicemeeter seem complicated to get either one to work.
strange mountain
Use Virtual Audio Cable lite instead. VB-Cable and Voicemeeter seem complicated ...
no i mean when i turn it on my speaker is mute and after exiting it goes back to normal
hallow thistle
no i mean when i turn it on my speaker is mute and after exiting it goes back to...
Is that a solved?
knotty moth
no i mean when i turn it on my speaker is mute and after exiting it goes back to...
more likely wrong settings, you should use headphones to listen (the monitor one)
or tell me if you want to show the screenshot
strange mountain
more likely wrong settings, you should use headphones to listen (the monitor one...
uhm, oh i think you guys got it wrong i mean it comes from voicemeeter i followed the instructions of Realism on Aihub then when it comes to choosing speakers when i choose the speaker i use to listen to the sound it loses all the sound to be able to listen.
hallow thistle
If you could explain it more clear though.
Otherwise, you can send your screenshot here. Words alone aren't enough to understand as picture.
knotty moth
uhm, oh i think you guys got it wrong i mean it comes from voicemeeter i followe...
check the audio routing settings in voicemeeter
strange mountain
Otherwise, you can send your screenshot here. Words alone aren't enough to under...
sure
strange mountain
Why are you deleting your messages?
wrong reply 💀
strange mountain
Why are you deleting your messages?
cant send screenshot
knotty moth
cant send screenshot
!give-media-perms 1h @strange mountain
knotty moth
that includes the settings in voice changer
strange mountain
that includes the settings in voice changer
it works fine but when i use voicemeeter there is no sound and the video says audio render error.
strange mountain
.
knotty moth
it works fine but when i use voicemeeter there is no sound and the video says au...
that's not the right answer
it is the voice changer screenshot
strange mountain
it is the voice changer screenshot
wait
this is ?
💀 I don't understand anything
strange mountain
voicemeeter is not a voice changer application
.
strange mountain
it should be mostly your real mic
noooooo ik, but i mean i followed the instructions to the output selection when i selected my headphone it immediately blocked the sound
hallow thistle
.
Oh wow. You didn't set a microphone on W-Okada, right? You seem confused yourself on this one.
strange mountain
Oh wow. You didn't set a microphone on W-Okada, right? You seem confused yoursel...
I just opened it and haven't set it up yet..
but it from my OutPut get mute when i use voicemeeter. it issue
knotty moth
<:skull_sob:1159372531992645662> noooooo ik, but i mean i followed the instructi...
voice changer will process the input audio from the setting
none means none will be passed to the output
hallow thistle
If input is set to none on W-Okada, then W-Okada would process nothing.
Please make your statement more clear. I don't think Voicemeeter would loop its audio into W-Okada by itself.
strange mountain
none means none will be passed to the output
NO
😭
my output
cant hear anything
hallow thistle
You think you know what you're doing?
strange mountain
You think you know what you're doing?
YES I KNOW
strange mountain
you mean the voice changer output?
no my headphone when use voicemeeter
knotty moth
no my headphone when use voicemeeter
you didn't even point where is the headphone in the settings
hallow thistle
I don't know. I don't use VB-Cable or Voicemeeter. I simply set my own W-Okada output to "Line 1 (Virtual Audio Cable)".
knotty moth
.
also all the audio settings in the voice changer is completely wrong
please re-read the guide
strange mountain
please re-read the guide
BRUH I read all but it from speaker output when use voicemeeter
knotty moth
BRUH I read all <:skull_sob:1159372531992645662> but it from speaker output when...
you should better draw and show the audio routing scheme you really want
strange mountain
you should better draw and show the audio routing scheme you really want
i cant explain thank you i will look into my headphone repair.
hallow thistle
On W-Okada, with server audio mode on, the input/output/monitor settings are supposed to look like this.
knotty moth
<:cat_seriously:1366692376575213599> i cant explain thank you i will look into m...
bruh your headphones not working when testing the sounds alone?
hallow thistle
Here's with client audio mode.
hallow thistle
bruh your headphones not working when testing the sounds alone?
Or it could be he set Voicemeeter to output into another but wrong speaker/headphone. 
strange mountain
Or it could be he set Voicemeeter to output into another but wrong speaker/headp...
NOOOOOOOOOOOOOOOOOOOOOO it make my headphone mute yep
hallow thistle
NOOOOOOOOOOOOOOOOOOOOOO <:skull_sob:1159372531992645662> it make my headphone mu...
No excuse.
This is the only speaker/headphone device my laptop has. Integrated sound card. The Line 1 is from Virtual Audio Cable lite, while Dell one is HDMI audio.
strange mountain
really no one understands my problem and i don't know how to explain
hallow thistle
If Voicemeeter being set to automated muting one of your headphones, find a way to disable it.
strange mountain
If Voicemeeter being set to automated muting one of your headphones, find a way ...
cant
knotty moth
really no one understands my problem and i don't know how to explain
ye you can blame your teammates for losing a game, until you see how good or bad you played it
hallow thistle
That's because you didn't explain in more clear sense, hence why I don't understand you. Sorry, but the issue about using Voicemeeter with W-Okada hits pretty much dead end now. Can't help with that if you continue to use that aforementioned program so.
You can either give up using Voicemeeter with W-Okada in favor of Virtual Audio Cable or continue to torture yourself using that program. 
pastel oak
On wokada
Input mic
Output Line 1
On voicemeeter
The screenshot correct, stereo input 1 line 1
On discord games obs
Input voicemeeter aux output
No point in routing it 2 ways though unless youre using voicemeeter for equalizing
I didnt read the question tho 
knotty moth
cant
well if you don't even know what to do, don't use voicemeeter and exactly follow the audio settings in the guide
strange mountain
well if you don't even know what to do, don't use voicemeeter and exactly follow...
ok ty
simple ore
for most people this version is a major overkill
potent bone
whats the difference between those three
full moss
done i downloaded it
echo meadow
@low shard
got no pic perms bruh
ive run on a problem in collab
using rvcaicovermakerUI
strange mountain
for most people this version is a major overkill
but i cant hear anything from my headphone 💀
viscid moss
i think its because, i run without gpu?
It's weird but I think so. I'm testing with GPU and it works
echo meadow
It's weird but I think so. I'm testing with GPU and it works
yeah, without gpu, this msg comes
brittle wing
Is Melroformer de-echo dereverb a good option or should I use anuew's dereverb & Uvr deecho?
Like...I wanna remove echo and reverb in one go
viscid moss
Is Melroformer de-echo dereverb a good option or should I use anuew's dereverb &...
Anvuew's one
brittle wing
Anvuew's one
Dereverb & deecho in One is by sucial.
simple ore
it was compiled with other library set, so yeah
joys of linux
it looks like the UI thing
may want to just build it
low mason
anyone know the newest version voice changer client oda for win ndivia force
latent kettle
anyone know the newest version voice changer client oda for win ndivia force
-realtime
latent kettle
The first one
low mason
tyty
full moss
can anyone help me
i try to talk but the software is not replacing the voice but i do hear it
sour venture
I'm trying to run the voice changer client but I do not know which one opens it
low mason
The first one
there 5 assets i need to download 5 of them right ?
brittle wing
Is overlap=8 enough?
olive plover
Force FP32 mode: on (THIS IS OFF BY DEFAULT!) Turning this on improves stability, significantly reduces glitching/artifacting, increases VRAM usage by 200 MB.
I put it ( on ) will it improve the sound quality?
random spindle
Exception: Invalid device_id argument supplied 1. device_id must be in range [0, 1).
Wtf?
RVC
odd isle
whats the difference between those three
did u figure it out?
potent bone
no
burnt thorn
full moss
how to turn on the voicechanger
full moss
what's the best rvc gui
can u help me
stray copper
can u help me
no
odd shale
and the most user friendly
You can actually either use Applio locally or on cloud.
-rvc
stray copper
i used to use applio
odd shale
i used to use applio
What's your GPU?
stray copper
RTX 4060 Ti 8GB
full moss
Then you'll be fine on local then.
can u
these are my settings
alright i gotta go
bye
stiff goblet
should i keep traning ? i'm at epoch 300 @simple ore @odd shale
low shard
these are my settings
#1371568582848417834 let's talk here
drowsy veldt
can someone help me rq? I have been using a google collab link for basic conversions titled "hina mod" (the one with a pop up anime girl picture) and i like this collab because the amount of conversion settings and options. anyway, for a month or 2 everytime i press "run cell" it says gradio live cant be used, or theres "ModuleNotFoundError: No module named 'gradio" what should i do? i try to avoid using anything other than collab if possible, ive tried weights.gg but it lacks versatility. if anyone has any idea please add me
low shard
can someone help me rq? I have been using a google collab link for basic convers...
Which Hina mod?
Send the link
Hina mod just means it's a modified version made by Hina which is one of our engineer
drowsy veldt
low shard
Also, tell your PC GPU
Because google colab is a cloud service meant only for bad PC users
low shard
https://colab.research.google.com/github/hinabl/AICoverGen-Colab/blob/main/Hina_...
Basically it's ourdated
What's your PC GPU?
There's a better tool for automatic ai cover generator
drowsy veldt
im fairly new to conversions and i havent had trouble in the past
im trying to find my gpu myb
silent stratus
is the applio lightning ai port back or nah
low shard
im trying to find my gpu myb
You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU
Here you can find your GPU easily if it's shy 
low shard
is the applio lightning ai port back or nah
Wdym
drowsy veldt
intel r UHD graphics is what it says (hopefully thats the right info)
low shard
the applio on lightning ai has been taken down for like a month now
Oh that, yes it's back https://docs.aihub.gg/rvc/cloud/applio-lightning-ai/
low shard
intel r UHD graphics is what it says (hopefully thats the right info)
Oh that's a weak integrated graphics, do you have any other GPUs?
silent stratus
Oh that, yes it's back https://docs.aihub.gg/rvc/cloud/applio-lightning-ai/
thankss
drowsy veldt
Oh that's a weak integrated graphics, do you have any other GPUs?
i dont think so bro, im just on my old laptop 
its never been a problem till nowwww
low shard
i dont think so bro, im just on my old laptop <:misc_cry:1176674698629750975> i...
Then use the cloud version of aicovermaker https://docs.aihub.gg/rvc/cloud/aicovermaker/
modest willow
There's a better tool for automatic ai cover generator
what's the tool? just booted up the old aicovergen and it doesn't even launch 😩
low shard
what's the tool? just booted up the old aicovergen and it doesn't even launch 😩
AICoverGen is dead, it's never coming back
modest willow
ripu
low shard
The new tool that makes automatic ai covers is aicovermaker
drowsy veldt
ill give it a try, i have no idea what ngrock or anything is bc im new but ill fugure it out. are there any more collab notebooks you know off of the top of your head that would run/leverages cloud based gpus? ive been asking chat gpt for assistance and it says it shouldnt be TOO much of a big deal if i have good internet
low shard
It's better anyways
modest willow
https://docs.aihub.gg/rvc/cloud/aicovermaker/#aicovermaker-kaggle this one works for local right?
low shard
https://docs.aihub.gg/rvc/cloud/aicovermaker/#aicovermaker-kaggle this one works...
This is for cloud, look into the local rvc part for local aicovermaker
What's ur PC GPU tho
drowsy veldt
AICoverGen is dead, it's never coming back
modest willow
rtx 5070 ti if your askin me
low shard
ill give it a try, i have no idea what ngrock or anything is bc im new but ill f...
The ones in the docs are good, u wouldn't really need other cloud computing services even because they aren't officially supported by us and would be paid
low shard
rtx 5070 ti if your askin me
Should work then
@viscid moss aicovermaker has Rtx 50 serie support, right?
drowsy veldt
The ones in the docs are good, u wouldn't really need other cloud computing serv...
im just trying to find something user friendly but also good with conversion settings like accent pitch and stuff aswell
modest willow
i found it, thanks man o7
low shard
i found it, thanks man o7
Yw and lmk
low shard
im just trying to find something user friendly but also good with conversion set...
There isn't like a 1 click option, the ones mentioned in the docs are the easiest and free ones, they are explained step by step
Also I think you're used to the google colab user interface
drowsy veldt
true ill try the link you suggested, thanks!
low shard
true ill try the link you suggested, thanks!
Yw and lmk
viscid moss
<@274566299349155851> aicovermaker has Rtx 50 serie support, right?
nope
just UVR5 UI
viscid moss
wwwooould it work if i put the covermaker* stuff on a mac instead?
it will def not work, it works only on NVIDIA PCs
modest willow
it will def not work, it works only on NVIDIA PCs
not to bug but any idea when the 50x will be supported? ballpark 
simple ore
pytorch for 5000 series with cu128 is available
modest willow
pytorch for 5000 series with cu128 is available
if i ran this inside the folder thru a cmd prompt would it update it? 
env\python -m pip install torch==2.7.0 torchvision==0.22.0 torchaudio==2.7.0 --upgrade --index-url https://download.pytorch.org/whl/cu128
also when training models, sometimes when it saves an epoch my pc shuts down ain't that fucked up
knotty moth
also when training models, sometimes when it saves an epoch my pc *shuts down* a...
what is the blue screen error?
if it's only abrupt shutdown, more likely PSU issue (transient spike/insufficient wattage) or faulty cpu/gpu
modest willow
if it's only abrupt shutdown, more likely PSU issue (transient spike/insufficien...
well the gpu is new and my psu is 750w uhh
knotty moth
well the gpu is new and my psu is 750w uhh
RTX 4070 with normal OC on 750w psu shouldn't have the issue
viscid moss
if i ran this inside the folder thru a cmd prompt would it update it? <a:pepeW:1...
this should work for rvcaicovermaker. I hope so
simple ore
well the gpu is new and my psu is 750w uhh
what gpu?
for 5070ti maybe too low
although the recommended is 700w by nvidia and around 750w by manufacturers
anyway, if the computer shuts down during high load, that's psu triggering overvoltage or something
viscid moss
Guys, we have a working realtime colab ?
Dietieris fork colab is disconnecting and hina's modified one has a problem with faiss
median plinth
Hello everyone, one question: what do you think is the best model, program, or website for separating voice and instruments (bass, drums, and others)?
I need to know
short storm
Idk but
I know a website you can generate a credit card or debit card for yourself
Validated
latent kettle
Hello everyone, one question: what do you think is the best model, program, or w...
Uvr 5 Ui
latent kettle
there 5 assets i need to download 5 of them right ?
What do you mean. There is only 1 zip file
hexed imp
Hey everyone!
Does anyone still have a ChatGPT Plus invite link for the free trial?
I’d love to test GPT-4o, but can’t afford the subscription right now.
I’d really appreciate it – thank you so much!
pastel oak
whats the difference between those three
Sorts voice models by the hz they were trained on
@odd isle since you asked too
pastel oak
Hello everyone, one question: what do you think is the best model, program, or w...
Easiest for free is mvsep.com, use bs roformer one of the newest
For drums bass etc one of the other models, dont remember name but it says what it separates when you select them
pastel oak
Guys, we have a working realtime colab ?
Dietieris fork colab is disconnecting ...
In that case no, need premium colab
pastel oak
how protocol : "rest" is better than "sio"
Sio and rest are communication methods for web applications, sio has less delay than rest
hazy eagle
Sio and rest are communication methods for web applications, sio has less delay ...
thanks but idc about delay, which has better quality?
pastel oak
thanks but idc about delay, which has better quality?
No difference, this is not important for quality
Fp32 mode, higher crossfade length, higher chunk, higher extra are relevant for quality
If you dont care about delay i assume you dont need realtime? You can record whatever you need and do inference in rvc like post production
full moss
can anybody help me my audio doesnt work
full moss
No difference, this is not important for quality
can you help me?
full moss
oh
still
doesnt work
i cant hear my voice + my ai voice
it doesnt detect my audio
simple ore
it doesnt detect my audio
client = using browser to send the audio over to server (usually local server anyway), server = using system's audio ins and outs
if you're using it on a local PC using server is a better option
input: your microphone, output: virtual cable, monitor: headphones if you want to hear your changed voice
forest verge
training a model using Replay and when testing, I've found that the 'S' sound in a model with only 5 epochs is a lot more natural and less 'lispy'/robotic compared to the same model with 250 epochs. the model is trained on 20 minutes of normal speech.
I wonder why that is
full moss
ok
full moss
client = using browser to send the audio over to server (usually local server an...
im using app
not web
and now im hearing it
but still
not changing my voicw
voic
e
and the perf is 300ç
and red
low shard
im using app
I replied to you to https://discord.com/channels/1159260121998827560/1371568582848417834
low shard
https://rentry.co/ForkVoiceChangerGuide#discord-games
Btw the link we are suggesting users now is the docs one
devout comet
why mi vc not working?
idk what im doing bad : (
low shard
-realtime
low shard
Which is the first link
low shard
why mi vc not working?
idk what im doing bad : (
Elaborate
What's your PC GPU? What tut link did you use? What is the issue?
devout comet
Elaborate
i download de vc client, i do everything and is not working.
im trying now the vc colab and send me an error
low shard
i download de vc client, i do everything and is not working.
im trying now the v...
Send the tut link you're following
devout comet
its spanish xd
low shard
its spanish xd
be sure to not use video tutorials for ai programs like Wokada
Send the link anyways just to check but 99% of the time they are just using an old original wokada version
Ai evolves at Sonic speed, videos can't keep up easily
devout comet
i think bcs is too old the videos and now have news versions
devout comet
Send the link anyways just to check but 99% of the time they are just using an o...
thats is what im trying now
i cant send u a screenshoot but the colab tell me that
ERROR: Could not find a version that satisfies the requirement faiss-gpu (from versions: none)
ERROR: No matching distribution found for faiss-gpu
Installing dependencies from requirements.txt...
ERROR: Ignored the following versions that require a different python version: 1.21.2 Requires-Python >=3.7,<3.11; 1.21.3 Requires-Python >=3.7,<3.11; 1.21.4 Requires-Python >=3.7,<3.11; 1.21.5 Requires-Python >=3.7,<3.11; 1.21.6 Requires-Python >=3.7,<3.11
ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu==1.13.1 (from versions: 1.15.0, 1.15.1, 1.16.0, 1.16.1, 1.16.2, 1.16.3, 1.17.0, 1.17.1, 1.18.0, 1.18.1, 1.19.0, 1.19.2, 1.20.0, 1.20.1, 1.20.2, 1.21.0, 1.21.1, 1.22.0)
ERROR: No matching distribution found for onnxruntime
Somthing like that
low shard
https://www.youtube.com/watch?v=BIu4FmSgyYk
Yup that's extremely outdated
devout comet
true
low shard
What's ur PC GPU
devout comet
i cant send u a screenshoot but the colab tell me that
ERROR: Could not find a ...
send me that
low shard
yes 🥲
Yup that's prehistorical nowadays, your only way is to either buy a better PC or use cloud (remote good PC) which is time limited on free tier
low shard
then i try the vc colab and is not working too
The colab is broken too
devout comet
oou
low shard
The "VC" is called wokada, there's 2 main versions, the original and the deiteris fork which is the best
Both versions are broken on Google colab
The only working cloud is wokada deiteris fork kaggle
devout comet
okey okey
low shard
okey okey
Kaggle is another cloud service which gives 30 hours weekly for free but needs a phone verification (so another google colab but advanced)
low shard
okey okey
https://docs.aihub.gg/rvc-voice-changer/cloud/w-okada-kaggle/ here's the Wokada deiteris fork kaggle link
devout comet
ill try
low shard
Just be sure to never use video tutorials for ai programs
Because they can't get easily edited like written guides
devout comet
https://docs.aihub.gg/rvc-voice-changer/cloud/w-okada-kaggle/ here's the Wokada ...
the guide of thats its in the website right?
that
low shard
the guide of thats its in the website right?
This is our ai hub official documentation, it has a guide for how to run the Kaggle version of wokada deiteris fork
All you need to do is read it up and you should be setup, for any issues let me know
devout comet
can u send me pls? : )
low shard
can u send me pls? : )
I already did
low shard
https://docs.aihub.gg/rvc-voice-changer/cloud/w-okada-kaggle/ here's the Wokada ...
That's the link
fast dagger
What's the best RVC today, is it still w-okada or have we evolved?
odd shale
What's the best RVC today, is it still w-okada or have we evolved?
RVC stands for retrieval voice conversion with prerecorded audios, not voice changer.
You're referring to voice changer, right?
fast dagger
ye
odd shale
ye
Welp, we mostly recommend deiteris fork, it's a fork of OG W-Okada with some improvements.
-realtime
odd shale
Read the first guide. (the one that says "most suggested")
fast dagger
Read the first guide. (the one that says "most suggested")
I will and then come back, ty
odd shale
I will and then come back, ty
You're welcome buddy.
low shard
RVC stands for retrieval voice conversion with prerecorded audios, not voice cha...
rvc does not stand for realtime voice changer 
odd shale
rvc does not stand for realtime voice changer <:boohooh:1230511774017650699>
Of course it doesn't XD
Maybe i didn't phrase my message correctly.
olive plover
Force FP32 mode: on (THIS IS OFF BY DEFAULT!) Turning this on improves stability, significantly reduces glitching/artifacting, increases VRAM usage by 200 MB.
I put it ( on ) will it improve the sound quality?
polar tendon
what's the tool? just booted up the old aicovergen and it doesn't even launch 😩
Use newer alternatives and eventually find one that works, a lot of those that i found were broken for some reason
polar tendon
For drums bass etc one of the other models, dont remember name but it says what ...
Demucs HT4 or something
worldly vigil
same error have you fix that bro?
latent kettle
Install Gradio
hexed imp
Hey everyone!
If anyone has one of those magical ChatGPT Plus free trial invites lying around, I’d be beyond grateful!
Really curious to explore GPT-4o, but my wallet’s on a break.
Thanks a ton in advance — you’re awesome!
simple ore
if it cant find gradio it means that most likely all other requirements did not get installed
brittle wing
brittle wing
rvc does not stand for realtime voice changer <:boohooh:1230511774017650699>
Retrieval voice conversion
modest willow
hm. it might be the factory overclock doing it then, since my cpu should be using a lower wattage in theory
or my psu is just old, its an evga 750 bronze from over a decade ago
before they used a bunch of shitty chinese capacitors and had taiwan stuff inside
simple ore
or my psu is just old, its an evga 750 bronze from over a decade ago
so it probably lost 10-15% of its wattage
modest willow
wait that's a thing?
simple ore
you're risking your hardware with 10 year old psu
simple ore
but generally yes, a power supply will drop its ability to supply power over time due to the capacitors aging and drying up.
so it would make sense to get something +100w extra to account for that
you dont want PSU running at 95% max power
olive plover
Force FP32 mode: on (THIS IS OFF BY DEFAULT!) Turning this on improves stability...
??
full moss
can anybody help me https://discord.com/channels/1159260121998827560/1371568582848417834
fringe snow
Does crepe need more than 5:30 of audio? everytime I try to train it I get the error
"Not enough data present in the training set. Perhaps you forgot to slice the audio files in preprocess?"
even though everything is already spliced into over a hundred wav files
I'm using the colab version
simple ore
Does crepe need more than 5:30 of audio? everytime I try to train it I get the e...
total number of samples / batch size needs to be > 3
it wont work if your slices are <0.5s or >9s
hot sierra
Hey it's been a while since i trained a model in RVC, i learnt that we can use pretrains now. i want to make a singing / talking model and have the datasets for it like 15 minutes or something. which pretrain model should i use? i installed Titan but im not sure if its the right one
hazy silo
anyone know how to fix it? applio works fine on my pc — 4070 / 32gb ram / ryzen 5800x
Using MRF HiFi-GAN vocoder
Traceback (most recent call last):
File "C:\codename-rvc-fork-3\env\lib\site-packages\gradio\queueing.py", line 625, in process_events
response = await route_utils.call_process_api(
File "C:\codename-rvc-fork-3\env\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api
output = await app.get_blocks().process_api(
File "C:\codename-rvc-fork-3\env\lib\site-packages\gradio\blocks.py", line 2137, in process_api
result = await self.call_function(
File "C:\codename-rvc-fork-3\env\lib\site-packages\gradio\blocks.py", line 1663, in call_function
prediction = await anyio.to_thread.run_sync( # type: ignore
File "C:\codename-rvc-fork-3\env\lib\site-packages\anyio\to_thread.py", line 56, in run_sync
return await get_async_backend().run_sync_in_worker_thread(
File "C:\codename-rvc-fork-3\env\lib\site-packages\anyio_backends_asyncio.py", line 2470, in run_sync_in_worker_thread
return await future
File "C:\codename-rvc-fork-3\env\lib\site-packages\anyio_backends_asyncio.py", line 967, in run
result = context.run(func, *args)
File "C:\codename-rvc-fork-3\env\lib\site-packages\gradio\utils.py", line 890, in wrapper
response = f(*args, **kwargs)
File "C:\codename-rvc-fork-3\tabs\inference\inference.py", line 1017, in enforce_terms
return run_infer_script(*args)
File "C:\codename-rvc-fork-3\core.py", line 180, in run_infer_script
infer_pipeline.convert_audio(
File "C:\codename-rvc-fork-3\rvc\infer\infer.py", line 251, in convert_audio
self.get_vc(model_path, sid)
File "C:\codename-rvc-fork-3\rvc\infer\infer.py", line 432, in get_vc
self.setup_network()
File "C:\codename-rvc-fork-3\rvc\infer\infer.py", line 476, in setup_network
self.net_g = Synthesizer(
File "C:\codename-rvc-fork-3\rvc\lib\algorithm\synthesizers.py", line 86, in init
self.dec = HiFiGANMRFGenerator(
File "C:\codename-rvc-fork-3\rvc\lib\algorithm\generators\hifigan_mrf.py", line 333, in init
torch.nn.Conv1d(channel, 1, kernel_size=7, stride=1, padding=3)
UnboundLocalError: local variable 'channel' referenced before assignment
odd shale
Hey it's been a while since i trained a model in RVC, i learnt that we can use p...
Actually just use either OG pretrain or KLM 4.9 one
OG is the best one for talking models and KLM can give you a bit better pitch range for singing
hot sierra
tysm!
torn bluff
is there any model that's related to changing my voice pitch?
like lower/higher pitch
simple ore
anyone know how to fix it? applio works fine on my pc — 4070 / 32gb ram / ryzen ...
you're trying to use MRF hifigan for some reason
and there's something wrong with the config
hazy silo
you're trying to use MRF hifigan for some reason
yeah, it was a model i trained with MRF HiFiGAN on applio
and it works fine on applio
glacial pollen
<@1239634084133601423> fix the extra tabs
which one?
simple ore
self.upsamples are one level too deep
glacial pollen
no, the vocoder I meant
simple ore
mrf hifigan
glacial pollen
ur_path\odename-rvc-fork-3\rvc\lib\algorithm\generators
there
Btw Noobies, thx for info
hazy silo
`` ur_path\odename-rvc-fork-3\rvc\lib\algorithm\generators ``
i tried adding it, still got the same error
then i did a fresh git clone of the updated version (idk, maybe i messed something up lol)
but same issue again
ended up just grabbing the one from applio and dropping it into your fork, and that fixed it
glacial pollen
huh
hold on, gonna compare, perhaps there's some more changes done to mrf I haven't seen or something up with merges
oh yeah, I think I see it
for some reason some of the latter formatting was fked up
In any case, it's fixed on repo.
and ye, you did good as they are the same both fork and applio.
torn bluff
is there any model that's related to changing my voice pitch? like lower/higher pitch
vivid hornet
What Ai voice changer is everyone using? I used to use a good one on my old PC but can no longer find the download for it
simple ore
-realtime
vivid hornet
-realtime
hmm yea I used to use Wokada Realtime Voice Changer had a video to walk me through it but I have no idea what their new github even means or what im even looking at in the Wokada Deiteris Fork github i dont see anything that looks like the file i had to click before
If i remember from before all I needed to do was click some kind of Cuda file to install the program
simple ore
just download, unzip, run the .exe
vivid hornet
download what exactly? Thats my issue I dont see anything here that even remotely looks like a download just code I dont understand
simple ore
golden walrus
https://rentry.co/forkvoicechangerguide
sir, may i ask if Spin embedder work with W-okada ?
simple ore
sir, may i ask if Spin embedder work with W-okada ?
not yet, old W-okada had an embedder choise, but uses fairseq, which is an old model
and new fork does not have an option to change the embedder yet
golden walrus
simple ore
@dr87 made Vonovox, not sure if he added that there either
golden walrus
thank you, and thank you guys for making awesome stuffs. i will train these and put it on hold
distant turtle
-colab
hallow thistle
download what exactly? Thats my issue I dont see anything here that even remotel...
On Detris' W-Okada page on AI Hub website, there are these.
mint goblet
currently running mainline thru kaggle; model training works fine but i dont see an index outputed anywhere. my steps were
upload dataset (into a new root folder called "dataset) -> proces data -> feature extraction -> train index
is it located somewhere other than Assets/Weights and logs/model_name that i havent checked yet? kinda new to all of this
slow wolf
hello good, I tried to open the start_http.bat file but I get an error and the program does not open, does anyone know how to fix it?
hallow thistle
hello good, I tried to open the start_http.bat file but I get an error and the p...
You're trying to use the original version of W-Okada, which has two batch files to run the program, and also outdated. What is your PC GPU?
There's a newer W-Okada to download and use, which doesn't need a batch file to run.
slow wolf
rtx 2060
simple ore
currently running mainline thru kaggle; model training works fine but i dont see...
in some versions if you do train index before training, it deletes the index
hallow thistle
rtx 2060
Download the newer W-Okada from this guide. https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows
mint goblet
in some versions if you do train index before training, it deletes the index
oh, gotcha
ill give it a try again once the model fully finishes training then
knotty moth
oh, gotcha
ill give it a try again once the model fully finishes training then
for index training that doesn't take forever for 20+ min dataset, try Applio or running locally (should be runnable on cpu mode once you already have preprocessed files)
mint goblet
for index training that doesn't take forever for 20+ min dataset, try Applio or ...
will the index be like
cross compatiable (idk the word) with the mainline trained model?
knotty moth
will the index be like
cross compatiable (idk the word) with the mainline traine...
you can try on local mainline
just for index training, it doesn't need a GPU
mint goblet
wait i think i found the issue
my dumb ahhhh had my dataset path set to datasets instead of dataset 💀
wait yeah that was my issue
holy shit
hallow thistle
latent kettle
thank you if anyone will help me
Maybe you can try pip install gradio. Idk it will work or not
simple ore
thank you if anyone will help me
As I told you yesterday, you're likely using an old outdated colab that has not been patches so none of the requirements are installed, that includes gradio
and a hundred of others
peak path
Hello guys
I used https://huggingface.co/spaces/r3gm/RVC_HFv2 and duplicate that into my profile.
But I've got multiple errors, and I can't fixed it.
Is there a new version for that?
Requested omegaconf from https://files.pythonhosted.org/packages/72/fe/f8d162aa059fb4f327fd75144dd69aa7e8acbb6d8d37013e4638c8490e0b/omegaconf-2.0.2-py3-none-any.whl (from -r /tmp/requirements.txt (line 3)) has invalid metadata: .* suffix can only be used with `==` or `!=` operators
PyYAML (>=5.1.*)
~~~~~~^
Please use pip<24.1 if you need to use this version.
I have my own trained models and if I want to convert voices, one of the good tools out there is
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI
I wonder to know is there a another option to allow me convert my voices via command line in Windows and Linux.
waxen pike
set extra to 2.0 and try other models
I'm having the same problem as you. I swallow half the words.
peak path
I don't need the default pth files in Replay.
Is it possible to disable downloading?
Download required files
We need to download some required files first.
This might take a bit.
Downloading D-f048k-TITAN-Medium.pth
because that’s outdated
what’s ur pc gpu? What u want to do?
low shard
I'm having the same problem as you. I swallow half the words.
wdym same problem as me? I was helping you, did you try other models? Also you need to do a bit of voice acting to make it sound better
peak path
what’s ur pc gpu? What u want to do?
I have my own models (pth) already.
I want to use my models to create voice-to-voice.
Is it possible to do it via Linux command line? Without GUI
low shard
I have my own models (`pth`) already.
I want to use my models to create `voice-t...
What’s your pc gpu?
there’s a local way with another rvc fork
peak path
Old
GTX 750Ti
peak path
I create my models in Colab.
low shard
too weak to run it locally on your pc
peak path
I don't use it in my PC
I want to use it on my Linux server.
low shard
I don't use it in my PC
I want to use it on my Linux server.
If your server has a bad gpu, it’s not worth it
AI requires alot of good hardware
GPU is the most important thing for running ai
does your linux server have a good gpu?
peak path
Yes.
In many videos I saw the process of generating mp3 files in very fast
low shard
Yes.
In many videos I saw the process of generating mp3 files in very fast
what’s the gpu you’re using for your linux server?
the speed depends on the gpu
ai takes alot of power
also, you shouldn’t check video tutorials for ai programs bc they get outdated easily, all rvc videos are pretty old
peak path
what’s the gpu you’re using for your linux server?
Quadro RTX 6000
low shard
Quadro RTX 6000
should work, there’s this cli rvc fork if u want https://github.com/blaisewf/rvc-cli
peak path
Let me check, tnx
low shard
Yw
simple ore
Let me check, tnx
you dont even need cli, all applio's function can be done using core.py
rvc cli is just a copy of it
peak path
also, you shouldn’t check video tutorials for ai programs bc they get outdated e...
Those packages runs on Docker and has predefined requirements.txt.
Why do these errors occur?
peak path
you dont even need cli, all applio's function can be done using core.py
applio? What is that?
Does it have a GitHub repository?
simple ore
low shard
you dont even need cli, all applio's function can be done using core.py
why isn't it specified in the applio github repo?
instead of making a copy could be better to just add CLI documentation
simple ore
rvc-cli has not been updated in like 6 months
low shard
rvc-cli has not been updated in like 6 months
yeah, which is why i said why not just archive it and add a CLI documentation for applio?
simple ore
asking a wrong person 🙂
low shard
asking a wrong person 🙂
should i ping blaise about this in IAHispano?
simple ore
maybe 🙂
low shard
i mean, i was just suggesting
simple ore
it is not really what we suggest to users, but it is an option
low shard
it is not really what we suggest to users, but it is an option
do people ever ask for cli in iahispano?
i see sometimes devs asking for it even though it's not really common
simple ore
not really, people ask how to run something like tts + voice change or use an API and blaise pretty much always says use core.py
but is happens rarely
low shard
but is happens rarely
mm yeah it's not really common
simple ore
if someone is down to that level of interaction with Applio it is more beneficial to just skip core.py completely and use the modules directly
but it can be used to just make a bat file and semi-automate something
like calling kokoro to generate tts with emotions and then core.py to change the voice
safe echo
hi whats the best virtual audio cable to route my voice through? thank you
safe echo
does anyone know why it wont load? ty
simple ore
vac lite 4.70
safe echo
i wanted to avoid that seeing as theres no official download page/licensing.
any others?
simple ore
tf
hallow thistle
does anyone know why it wont load? ty
I saw what you deleted earlier. Which W-Okada version are you trying to use?
simple ore
i wanted to avoid that seeing as theres no official download page/licensing.
what are you smoking? https://vac.muzychenko.net/en/download.htm
hallow thistle
Old version of fork. Make sure you download the latest one from AI Hub website. https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows
safe echo
what are you smoking? https://vac.muzychenko.net/en/download.htm
ah sorry mate, the guide im using instant downloaded it without taking me the site so i didn't want to run it
safe echo
Old version of fork. Make sure you download the latest one from AI Hub website. ...
thank you, ill look
hallow thistle
The latest version is b2332.
safe echo
which version is best for RTX 4070 ti super?
i guess its just the b2332 ver? (ill just use that)
golden walrus
anyone know how to use vonovox from dr87 ?
hallow thistle
i guess its just the b2332 ver? (ill just use that)
Fork W-Okada would work with any GPU, regardless of any version. But I'd say to stick with latest one.
safe echo
i have a voice model in my old version, where do i find it to move it to my new version please?
hallow thistle
i have a voice model in my old version, where do i find it to move it to my new ...
model_dir
safe echo
you're a angel thank you 
robust wave
WHAT CAN DO THIS MODEL ?
upbeat carbon
can anyone help me?
grim bay
What does this means? Should I change this?
upbeat carbon
what in the world is that?
grim bay
what in the world is that?
It's below the Pitch Extractor setting in Google Colab RVCAICoverMakerUI
upbeat carbon
pitch extractor? wait so you can make it sound just like the character that you want?
grim bay
pitch extractor? wait so you can make it sound just like the character that you ...
If you have the character's model
halcyon panther
im annoyed that no matter what i do, i cant have stable voice on okada anymore
i have a 4070 with a i7 14700hx and i just dont
glacial pollen
ah sorry mate, the guide im using instant downloaded it without taking me the si...
You can always just.. you know, search for any potential original / author website, dl from there and compare hash between em
btw.
glacial pollen
What does this means? Should I change this?
Also, no, ideally you wanna keep it as contentvec
other ones are " more suited " for that given language, but you gonna pay the price of the model struggling with the languages different than the one in the embedder's name
ContentVec ( at least for now, until the spin situation is clear. ) is your best and most universal most-languages-friendly option
glacial pollen
More details the better
don't just post errors or vague commandline screenshots, always provide what you use and potentially, if the situation requries it, your hardware etc.
vale osprey
-colab
surreal latch
Hello, I wanted to make a model of a character, but it says that the dataset is very small (the character doesn't speak that much), but I separated the audios into 4 parts but it still gives that error
spring orchid
-colab
spring orchid
-rvc
vivid hornet
anyone know why my ping in Realtime voice changer skyrockets when in a game? it goes to like 38000+ ping then freezes
but as soon as I toggle out of the window it goes back to 1-2ms
safe echo
hey, is there someway to hear myself to test?
vivid hornet
hey, is there someway to hear myself to test?
Set Monitor under output to your speakers/headset that you use
glacial pollen
hey, is there someway to hear myself to test?
you can just route voice changer's output to vac
and then playback the vac
safe echo
and then playback the vac
Thank you!
glacial pollen
No problem ✨
safe echo
please does anyone have advice on how i can stop voices going "croaky?"
glacial pollen
please does anyone have advice on how i can stop voices going "croaky?"
oh, that can be cause of A) you got out of model's comfortable range ( i.e. you spoke / sung something too low or too high ) or B) you moved your head too much sideways relatively to the mic and stuff got, nicely put, distorted
but B is relatively rare so, must be the former
But then, maybe some real-time spec can answer more on some stuff ( I barely use it
analog obsidian
please does anyone have advice on how i can stop voices going "croaky?"
happens when models were not trained using raspy audios
it tries to mimic ur raspiness but it fails
glacial pollen
oh, or perhaps that
or maybe whispering
safe echo
ahhh got it, thanks for the advice you guys are extremely helpful ❤️
bit hard to get a males voice to female.
vivid hornet
Anyone know how to fix ping issues on the voice changer when playing a game?
I have it set to the proper parameters recommended by the guide but when I tab into Rust the ping skyrockets
safe echo
wasn't there a shop here somewhere to buy paid voices?
analog obsidian
yup admins removed it alongside model master role
safe echo
ah, right, unsure if im allowed to ask where to find more?
i did buy in the past.
i know theres the free stuff
glacial pollen
analog obsidian
honestly i have no idea if they allow commissions here but considering they also removed request models channel they may not like that happening again
but yea use #1175430844685484042 to find models
glacial pollen
alternatively you can just commission someone
safe echo
fair, thank you 🫡
glacial pollen
And as for weights.gg / .com , unsure
crude flame
honestly i have no idea if they allow commissions here but considering they also...
Actions will be taken against anyone who promotes or offers commission-based work in DMs, server profiles, or the server itself
analog obsidian
rip
glacial pollen
yea, that's a lil sad ngl
lucid creek
hello there there way use GPT-SoVITS on cloud { kaggle} without ui
analog obsidian
yea, that's a lil sad ngl
have you tried using a l1 mel applio model in realtime?
bc i think i found something
the model just does this (the fumiama one doesn't do that)
in the deiteris fork actually sounds very bad, it cannot speak properly at all
inference in applio works good
glacial pollen
have you tried using a l1 mel applio model in realtime?
bc i think i found somet...
I mean
I tested maybe like 3 or 4? models from applio and perhaps 2 from fumiama / mainline in voice changers
never had any issues like that
analog obsidian
have you tried using a l1 mel applio model in realtime?
bc i think i found somet...
this is L1, the code noobies ported from mainline
try using one
glacial pollen
I mean, there is no way something like this would result from how one ported / implemented the L1
L1 works, if it was jammed you wouldn't be able to train properly / and if you were, you'd see jibberish mumble in specs in tensor
analog obsidian
could be the f0?
glacial pollen
hard to say.. need to see the spec, wait
glacial pollen
as you can see
even on extended log scaling
it doesn't look right
harmonics are jammed
and mel scale
try to boost the transpose in rt vc
see what happens
analog obsidian
ok 1 sec im opening fumi's realtime again
analog obsidian
ok
glacial pollen
and now try to keep it at 0
but raise your voice's pitch
by half, if you can
or at least by 2/3rd compared to prev attempts
safe echo
can you download voices from weights.gg? seems to be just samples
analog obsidian
by half, if you can
i tried mainline inference and it worked fine
analog obsidian
can you download voices from weights.gg? seems to be just samples
u have to log in
safe echo
u have to log in
i did but it only says create
glacial pollen
i tried mainline inference and it worked fine
" i tried mainline inference and it worked fine " wdym
analog obsidian
" i tried mainline inference and it worked fine " wdym
watashi no ko i wa
no electric sound
glacial pollen
I mean, we talking about harmonics now, of the " weird " model
safe echo
OH found it, sorry 
brittle wing
i need help
analog obsidian
I mean, we talking about harmonics now, of the " weird " model
im talking about the buzzing sound that happened at the start of the first audio
glacial pollen
im talking about the buzzing sound that happened at the start of the first audio
oh, smh
brittle wing
on the hispano notebook
glacial pollen
why you didn't say
Lyery
brittle wing
on kaggle
analog obsidian
why you didn't say
Lyery
analog obsidian
on the hispano notebook
iirc applio's kaggle is broken atm
brittle wing
iirc applio's kaggle is broken atm
shit what do i use then
analog obsidian
shit what do i use then
i have no idea
analog obsidian
ic, how can i prevent the model from doing that? the mainline one works just fine
in realtime
like not doing that electric buzz at all
analog obsidian
<@775545133448953856> be so deadass do u genuinely not know where i could train ...
😭 im fr
glacial pollen
and if you get rid of it
brittle wing
i train locally
damn
brittle wing
Just train locally if you can
why do u think im cloud training lmao my laptop is not good enough
analog obsidian
hold on i think i know how to replicate that electric sound
brittle wing
then what is it
brittle wing
cause i was told to upgrade the ram by a staff member here
analog obsidian
hold on i think i know how to replicate that electric sound
ok so magically it's not doing that anymore
😭
glacial pollen
cause i was told to upgrade the ram by a staff member here
How about we first get to know your hardware
what's ur gpu and it's vram
glacial pollen
ok so magically it's not doing that anymore
😭
oh, can be your cable then / mice
some interference or bent cable or some signal from around, wifi etc
lots of variables
or perhaps your cable is around something that's charged up ( or next to speakers etc. you get the idea
Or you can try cleaning the port
safe echo
its so odd, any voice model i download sounds nothing like what the samples do.
analog obsidian
its so odd, any voice model i download sounds nothing like what the samples do.
match the pitch
or use index
brittle wing
what's ur gpu and it's vram
CPU: AMD Ryzen 5 7535HS w Radeon Graphics, 3301 Mhz, 6 cores, 12 logic idfk
GPU: RTX 2070 and it says its got like 4 gigs of dedicated video memory idk what tf is that idk if its vram
RAM: 8 GB
glacial pollen
its so odd, any voice model i download sounds nothing like what the samples do.
you gotta match the target pitch of the model / modulate / impersonate sometimes
also index on ( if you can adapt ) or off if you can't, else artifacts can occur if the model is mid or low quality
glacial pollen
CPU: AMD Ryzen 5 7535HS w Radeon Graphics, 3301 Mhz, 6 cores, 12 logic idfk
GPU:...
I mean...
there is not 4 gig rtx 2070
🧐
safe echo
you gotta match the target pitch of the model / modulate / impersonate sometimes...
its either robotic or like screeching grrr
what should my chunk/extras be at?
glacial pollen
its either robotic or like screeching grrr
well yeah, then you gotta be aware of 3 possible causes:
- Bad model
- Bad usage ( not adapting or matching the model's pitch / timbre
- Bad settings ( transpose n stuff
brittle wing
there is not 4 gig rtx 2070
idk bro lol i dont know about pc specs and shit
glacial pollen
what should my chunk/extras be at?
All is there, on the site
should be more or less trustworthy
glacial pollen
idk bro lol i dont know about pc specs and shit
well then I'll go and simplify it for you.
Yes your hardware is viable for training
( but in some harsher cases, you'd need to use ' checkpointing ' feature which gonna let you train on bigger batch size if needed but that'll slow down the training. )
Nevertheless, ye. you're good to go
brittle wing
well then I'll go and simplify it for you.
Yes your hardware is viable for train...
damn alr thank you
glacial pollen
damn alr thank you
yea yw.
Applio or my Fork, those are your options
but I assume you're not advanced enough so, go for applio ( fork's for advanced users
brittle wing
yea yw.
Applio or my Fork, those are your options
but I assume you're not advanc...
yeah i use applio normally
brittle wing
i've trained some models but all on notebooks
brittle wing
good luck then
thank you can i ping u if i face any other problem?
glacial pollen
As long I'll be free and it ain't something too complex ( for newbies ) to explain, sure
will try to help
analog obsidian
@glacial pollen ok i've tried again and im still having those buzzing and the model sometimes sound veryyy low/muffled
i know how to replicate that
i have to say Oh
mainline model doesn't do that at all
no interference going on here
glacial pollen
<@1239634084133601423> ok i've tried again and im still having those buzzing and...
means model's jammed
crude flame
i've trained some models but all on notebooks
lightning.ai is another good place for training if youre willing to go through the process of signing up. I havent had any issues with it so far. Only real downside is that you dont get a lot of gpu hours depending on what gpu you choose.
brittle wing
lightning.ai is another good place for training if youre willing to go through t...
i might try that fs depending on how this goes on my laptop thank you
glacial pollen
nothing beats the freedom of local training tbh
crude flame
nothing beats the freedom of local training tbh
fr
analog obsidian
means model's jammed
rip
welp im glad i havent deleted fumiama and now i have proper logs thanks to you 
glacial pollen
well, I'll try to train my own " og pretrains " so
Gonna see if I can beat any other " og like " ones oof
brittle wing
what settings do u recommend @glacial pollen
glacial pollen
there is no recommendation
it is case / set dependent
tho, I might have some ' basic ' settings for you to tweak further
what is your dataset's length?
brittle wing
it is case / set dependent
my set is like 15 mins long
brittle wing
it has like minimal silence
glacial pollen
Alright, in that case, for 15 mins..
is it " diverse " in content?
brittle wing
like i trimmed some of the silence
glacial pollen
or rather monotone? ( both patterns ( words, phonetics etc ) and pitch wise ( voice modulates a lot with their pitch or not?
brittle wing
it's raw vocals from a song so i'd say like in between prolly
not too monotone and there are some parts where he changes the tone on that song
glacial pollen
oh yeah, then try batch 8 as a safe starter point
brittle wing
bet
glacial pollen
then, if you're willing to go through it ( and / or if you're not safistified with what you get
you can try 6 or 4
( and if you get OOM / Memory issues / cuda memory issues etc etc. Try lowering the batch from 8 to 7 or 6 ( try each one consecutively in case of issues.
(( or yeah, enable checkpointing, that will also work fine.
brittle wing
you can try 6 or 4
( and if you get OOM / Memory issues / cuda memory issues etc...
okay
i'll have to stop training tho for that right?
glacial pollen
I mean, if you wanna retry with what I said yeah
( as in, you can't change those mid-training
brittle wing
alr bet
glacial pollen
just dl the precompiled one yea ( 3.2.9
brittle wing
also i confused my pc rtx with the laptop one smh that's mb it's not a 2070 its a 2050 and it is 4 gigs
i just double checked cause it seemed weird to me that this happened
idk if thats too big of a difference
glacial pollen
oh, well
then that does change a lot
now that you say it's 2050
oof
well.. guess you could still try using it, but surely not without " checkpointing " enabled
brittle wing
alr thank you
glacial pollen
brittle wing
mb i confused them but i dont really want to use my pc
glacial pollen
this or however it's described in applio
glacial pollen
mb i confused them but i dont really want to use my pc
In any case, don't expect much from 4 gig 2050, that's for sure. And yeah, got the link
one sec
brittle wing
thank you
glacial pollen
yw, again, best of luck ( you'll need it
brittle wing
In any case, don't expect much from 4 gig 2050, that's for sure. And yeah, got t...
alr thanks
thanks for the patience lol
glacial pollen
yeah np man, at least you're willing to go on with it
some just drop the convo too quick or give up mid-work, wasting my time lol
so I appreciate that.
junior gull
Soryr new here, and still learning but....Does anyone have the actual hifigan.pt -trained vocoder (not raw) to share? The repo it used to be on is dead....its not in any of the beta builds for RVC v2. And I can't seem to find it ANYWHERE. It's litterally the last thing stopping me from finishing my first voice model Please help!
glacial pollen
Soryr new here, and still learning but....Does anyone have the actual **hifigan...
oh?
wdym
there is no hifigan.pt
the is pretrained Generator and pretrained Discriminator
Both can be found on
there ^
f0G48k and f0D48k
( if you go for non f0 variants, you get ones without nsf hifigan support. )
brittle wing
some just drop the convo too quick or give up mid-work, wasting my time lol
thank you
glacial pollen
And if you'll need current mirror:
https://huggingface.co/IAHispano/Applio/tree/main/Resources
That's the one applio uses
( but if it's about hifigan and pretrained models, it's all the same 1:1. Both on rvc's repo and applio's
That should be all ye?
junior gull
i've spent three days building my voice model and combing out bugs..and I finally got it to train and now its hanging on building the index. I've ran the erros through AI to help identify and resolve...and it keeps insisiting its becuase the method i'm using iwth RVC v2 is dependent on Hifi-gan but in the rvc_infer.py when it calls for it it even says I need to "obtain this file myself" and the AI is telling me the same thing...the closest thing I could find was an untrained version of it I'm very new to this so if there is another way to do it, that doesn't invovle scrapping all my current work I'd love to try.
Reading the replies now!
glacial pollen
i've spent three days building my voice model and combing out bugs..and I finall...
I mean, first of all
you should provide info on what you were doing in the first place
lemme know what you used:
- training using pretrained models or not?
- which vocoder? ( hifigan, refinegan, mrf-hifigan
glacial pollen
Because I can already tell you, by stock, you're only provided with hifigan pretrained models ( for sample rates: 32, 40 and 48khz )
and refinegan / mrf-hifigan need their own
in terms of refinegan, you should ask @ noobies5663
junior gull
lemme know what you used:
- training using pretrained models or not?
- which voc...
sure....I started from scratch
trained using 1000 epochs, a roughly 15 min sample
glacial pollen
..
junior gull
it took 4 hours >_<
glacial pollen
you absolutely should not do it
junior gull
it sounds sooo clean so far though in the chopped up samples
glacial pollen
a lot of models in fact already top in performance by 100-200, sometimes 300 epochs
but then, it depends ig, there are some uhm.. " special " cases
glacial pollen
Well, in any case
you don't go by " I'll train for N epochs " mindset
instead, you run the training and actively monitor the model's performance using tensorboard
In other words, you train for as long as the model needs
junior gull
i just figured for my "first" one i would max the detail and spend the extra time to get it right was my mindset
i just had it run over night
glacial pollen
I thought so too, initially, dw ( when I first started learning it 2 years ago or so )
it happens
junior gull
lol
glacial pollen
In any case, tensorboard is what you need my guy
junior gull
ah ...so you can examine the samples as the come out and when it gets to a poitn you like you can stop
glacial pollen
I mean yeah
but if you want most accurate loggings / metrics
My fork's ideal for that, also you can easily use your own " inference examples " samples ( for tensorboard evaluation, easily )
But if you don't care about ' pin point ' accuracy in metrics then you're 100% fine with APplio
junior gull
right now I'm trying to compile all those pieces with the index to make the completed voice model
glacial pollen
Sure
junior gull
the webUI kept hanging so I've been doing it from the console
glacial pollen
In any case, creating index should take no more than 5-20 seconds? ( in majority of cases
junior gull
is there a specific file or error i can grab tha twill better explain my problem?
junior gull
kk sec
glacial pollen
( ps. and I do deeply hope you are not using original rvc.
junior gull
i don thtink so? i slected the newest non beta version and in the web ui i it was saying i was usinv RVC v2 as well
magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '\x13'.
thats' where the process dies
glacial pollen
i don thtink so? i slected the newest non beta version and in the web ui i it wa...
well, that is rvc
why are you not using Applio?
glacial pollen
Well, in that case you should switch
as I only help with issues that result from using Applio / Fork
and as for the error, that's alien to me
junior gull
glacial pollen
https://huggingface.co/spaces/r3gm/RVC_HF
this one?
no
Now you need to make a choice, do you want to use Applio or the Fork
( aka, Advanced features and precise logging vs more user-friendly with less precise logging
then I'll link you the proper one
junior gull
I'd rather go the route that offers the higher quality result even if it takes more trouble.
Bonus points it its something I can use with ComfyUI....just learning how to use that too
brittle wing
@glacial pollen brooo
fallow solstice
yo im making a AI anybody wana help out dm me i guess its chill doing it with some friends but could use some help
glacial pollen
<@1239634084133601423> brooo
welp, at this point idk. guess your laptop's cooked
maybe can't handle it
glacial pollen
I'd rather go the route that offers the higher quality result even if it takes m...
Well, from technical stand point, both should more or less offer the same end results
however as I said, if one wants to tweak a lot, have more user-friendly and accurate logging etc, my fork's the way to go
( quick lr changing, other optimizers, few extras from me and so on. )
if one doesn't need that? then Applio
that's all there really is to it