Hi. I have a problem. Before I made format of my computer RVC worked amazing. But after format, when I try voice inference, the quality is awful even with the same settings I had ealier. Any idea?
#✨│ai-help
fresh geode
inner skiff
-realtime
azure marshBOT
-realtime
This interaction has expired, use the command
/guides realtimeif you wish to see it again.
proper shale
It isn't installed like an application (at least the main forks out there) so yeah, no
Kinda weird tho 
violet heron
what colab do you recommend?
For inference or training?
keen garden
uhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
yes
srry i know nothing about this
i just wanna make those ai covers
what should i use
violet heron
i just wanna make those ai covers
I would recommend using a HuggingFace space instead for covers
-spaces
keen garden
k
violet heron
Try Ilaria RVC zero
hybrid bolt
-colabs
brittle wing
help when i try to download models it asks me for a password
rare gobletBOT
Ayo? @brittle wing level 2 !!! 
hybrid bolt
what colab should i use for just making ai covers? RVC mainline or AICoverGen
loud oracle
can someone plz help me my files dont work on voice ai like they wont importr
low shard
can someone plz help me my files dont work on voice ai like they wont importr
we dont offer help for voice.ai, we use wokada here
low shard
help when i try to download models it asks me for a password
what are you using?
analog obsidian
did i do something wrong
F0 Est.: use rmvpe, not crepe
Audio in: your real microphone
audio out: Virtual audio cable (Line 1)
mon: none
brittle wing
yall help with mangio?
seperation doesnt work for me
at all
no seperation no reverb or echo
brittle wing
F0 Est.: use rmvpe, not crepe
Audio in: your real microphone
audio out: Virtual ...
okay let me try
queen sonnet
I created in rvc collab
/content/AICoverGen-NO-UI-en/song_output/ikSV1DYTxXU/GTA SA - OG Loc Rap [Music Video] (Am Ver).wav
What is this
brittle wing
F0 Est.: use rmvpe, not crepe
Audio in: your real microphone
audio out: Virtual ...
this worked but the thing is
its kinda choppy and i can hear myself twice
analog obsidian
its kinda choppy and i can hear myself twice
enable echo and sup2
brittle wing
no seperation no reverb or echo
?
analog obsidian
its kinda choppy and i can hear myself twice
play with those settings too until it works fine
less chunk delay = more gpu usage
more extra = more cpu usage
brittle wing
alright ill try
brittle wing
wdym by this, inference is not working?
can you dm ill show you cause i cant really explain it
analog obsidian
ooooh i think you're using uvr inside mangio right?
brittle wing
noo
analog obsidian
i believe you can post pics here, so if you need to send one do it
analog obsidian
i dont really get it tbh
remove spaces from audio name
brittle wing
okay ill try
brittle wing
ty honestly didnt know that
analog obsidian
brittle wing
does that also apply to training because mine just doesnt load it just sits there
analog obsidian
does that also apply to training because mine just doesnt load it just sits ther...
same
remove spaces from any file u want to use in rvc
and any special characters
brittle wing
solves alot actually
rare gobletBOT
Ayo? @brittle wing level 2 !!!
analog obsidian
okay so now on seperation its been processing for 3 mins
what pitch extraction method u used?
and how long is the audio in minutes
brittle wing
what pitch extraction method u used?
VR-DeEchoDeReverb
analog obsidian
no i mean, in rvc
brittle wing
im not training it
analog obsidian
... are u using uvr rn?
brittle wing
no
analog obsidian
ok let me explain this simple
so you are in the inference tab
loaded your model
analog obsidian
don't use that, its bugged
brittle wing
don't use that, its bugged
oh
analog obsidian
instead use uvr itself
brittle wing
instead use uvr itself
ill check that
analog obsidian
its the same thing but standalone and not bugged
marble forge
auto silence remover with uvr?
is that a thing?
like it cuts out the silence in audios
analog obsidian
like it cuts out the silence in audios
audio labels does this
what it does is skipping the silences and only selects the actual audio
then you're able to export every audio in different files
frozen harbor
I need help with setting up a voice AI for Goku from Dragon Ball. Can someone help me, please?
violet heron
I need help with setting up a voice AI for Goku from Dragon Ball. Can someone he...
What do you need help with
frozen harbor
@violet heron I'm looking to record for a podcast, but have the recording in Goku's voice. I was wondering if that would be possible during a recording, or would I have to record my vocals, then use the recording to turn it into a Goku AI voice over
violet heron
<@858923165379199007> I'm looking to record for a podcast, but have the recordin...
You can do it in real time or using a audio file
But it sounds better when you convert it using a audio file
To convert a audio file using a voice of your choice use RVC
frozen harbor
@violet heron how would I do that? Do I need to pay for that it is it free?
rare gobletBOT
Ayo? @frozen harbor level 1 !!!
violet heron
<@858923165379199007> how would I do that? Do I need to pay for that it is it fr...
Free and open source
violet heron
To convert a audio file using a voice of your choice use RVC
RVC can run locally or on a HuggingFace space
frozen harbor
@violet heron I have the Goku AI model downloaded. But idk where to go from there 🤔
violet heron
<@858923165379199007> I have the Goku AI model downloaded. But idk where to go f...
That is why I’m telling you how to use it
frozen harbor
@violet heron my apologies
violet heron
Once again, to convert a audio file to a voice model of your choice you would use RVC, which is a FOSS piece of software meant for cloning voices. Most if not all models in #1175430844685484042 are compatible with it.
RVC can run locally on your computers hardware or via a server hosted on a HuggingFace Space
It’s recommended to use a HuggingFace space instead of running it locally if you have poor hardware (for ex no GPU)
-spaces I recommend Ilaria RVC Zero if your gonna use a HuggingFace space
violet heron
-local here are some options that run locally on your computers hardware, it is recommended to have a decent GPU so it doesn’t take long to convert a file.
azure marshBOT
-local here are some options that run locally on your computers hardware, it is ...
violet heron
Applio has some cool extra features and has simpler ui compared to original RVC
But Original RVC is the least bloated but has the bare necessities
Tho some parts aren’t fully translated
Most cases it’s unzipping and opening a .bat file and waiting for the gradio ui to load in your browser
brave swallow
have anyone did xtts fine-tuning I didn't get good results
frozen harbor
Applio has some cool extra features and has simpler ui compared to original RVC
I keep getting an error when trying to open a project on the website
sonic agate
titan, ov2 or klm 4.1?
marble forge
RuntimeError: The expanded size of the tensor (12800) must match the existing size (0) at non-singleton dimension 1. Target sizes: [1, 12800]. Tensor sizes: [0]
anyone know how to fix?
proud elbow
RuntimeError: The expanded size of the tensor (12800) must match the existing si...
seems like mismatched sample rate and shit from previous preprocessing & feature extraction
marble forge
seems like mismatched sample rate and shit from previous preprocessing & feature...
how do i fix it?
proud elbow
how do i fix it?
start over from preprocessing, obviously 
marble forge
start over from preprocessing, obviously <:PES2_SusGe:838743169822752778>
so i have to restart the entire training?
proud elbow
so i have to restart the entire training?
start over or stick to the sample rate you have chosen before
marble forge
its the exact same sample rate nothing changed
proud elbow
I have no much idea, what rvc are u using?
oh well, how is the dataset format? a dataset folder containing wav files shouldn't throw random issues unless corrupted
marble forge
they are mp3 files and i think the reason that happened was because i removed two of the dataset mp3s even though there was 127 but i figured out a way to start the training from the same spot the old one left off and i think it might work the same but im not sure:
@proud elbow
what do you think
proud elbow
they are mp3 files and i think the reason that happened was because i removed tw...
changing dataset means you have to start over, otherwise it would still use the already preprocessed dataset before
torn flare
[VCClient] Access http://127.0.0.1:18888/
[VCClient] wait web server...0 http://127.0.0.1:18888/
[VCClient] wait web server... done 200
[2024-06-26 05:58:06] connet sid : cLGA42_pZ7oH1VhxAAAC
[2024-06-26 05:58:06] connet sid : v51mMX0vsl_jAFWYAAAD
paramDict {'voiceChangerType': 'RVC', 'slot': 4, 'isSampleMode': False, 'sampleId': None, 'files': [{'name': 'GawrGura_e140_s18620.pth', 'kind': 'rvcModel', 'dir': ''}, {'name': 'added_IVF4936_Flat_nprobe_1_GawrGura_v2.index', 'kind': 'rvcIndex', 'dir': ''}], 'params': {}}
RVC:: slotInfo.modelFile GawrGura_e140_s18620.pth
[Voice Changer] Official Model(pyTorch) : v2
SlotInfo::: RVCModelSlot(slotIndex=-1, voiceChangerType='RVC', name='GawrGura_e140_s18620', description='', credit='', termsOfUseUrl='', iconFile='', speakers={0: 'target'}, modelFile='GawrGura_e140_s18620.pth', indexFile='added_IVF4936_Flat_nprobe_1_GawrGura_v2.index', defaultTune=0, defaultIndexRatio=0, defaultProtect=0.5, isONNX=False, modelType='pyTorchRVCv2', samplingRate=40000, f0=True, embChannels=768, embOutputLayer=12, useFinalProj=False, deprecated=False, embedder='hubert_base', sampleId='', version='v2')
[Voice Changer] update configuration: modelSlotIndex 1719377956004
gin_channels: 256 self.spk_embed_dim: 109
[Voice Changer] generate new embedder. (no embedder)
[Voice Changer] Loading index...
Try loading... model_dir\4\added_IVF4936_Flat_nprobe_1_GawrGura_v2.index
i just get this for some reason
and is stuck on that
marble forge
changing dataset means you have to start over, otherwise it would still use the ...
oh ok
torn flare
i litteraly reinstalled and still have this [VCClient] Access http://127.0.0.1:18888/
[VCClient] wait web server...0 http://127.0.0.1:18888/
[VCClient] wait web server... done 200
[2024-06-26 06:20:07] connet sid : PZB4l2E2K0dAjpw9AAAC
[2024-06-26 06:20:07] connet sid : B7BMx7LW08PnUIBkAAAD
[Voice Changer] update configuration: modelSlotIndex 1719379231001
[Voice Changer] generate new embedder. (no embedder)
[Voice Changer] Loading index...
[Voice Changer] Index file is not found
rare gobletBOT
Ayo? @torn flare level 1 !!!
torn flare
cant even use the preinstalled models
violet heron
I keep getting an error when trying to open a project on the website
What’s the error?
frozen harbor
I believe my laptop cant handle the default settings it's presenting me with
violet heron
I believe my laptop cant handle the default settings it's presenting me with
Which link are you using?
And what is the error
torn flare
either get generating sola buffer or index file not found
violet heron
either get generating sola buffer or index file not found
did the model come with a index in the zip?
If not you can’t do much about it, if the index isn’t in there the accent is most likely baked in anyways
frozen harbor
@violet heron Do i drag and drop the php file to use the voice i want?
violet heron
<@858923165379199007> Do i drag and drop the php file to use the voice i want?
RVC doesn’t use php file do you mean pth?
Depending on the fork you drag and drop in the model loader or downloader
frozen harbor
RVC doesn’t use php file do you mean pth?
I'm trying to use Goku's voice and select it here
violet heron
I'm trying to use Goku's voice and select it here
I would recommend using Ilaria RVC over the one your using
Since that one uses a cpu for its server which is extremelyyyyy slow
frozen harbor
Ok, ill click the link now. Thank you
violet heron
frozen harbor
https://huggingface.co/spaces/TheStinger/Ilaria_RVC
So the Goku voice they have is the japanese version. Is there a way i can use the one i downloaded?
rare gobletBOT
Ayo? @frozen harbor level 2 !!!
violet heron
So the Goku voice they have is the japanese version. Is there a way i can use th...
Yes, model loader
frozen harbor
Yes, model loader
Index i'm guessing is for the script i want the character to say? or can i skip that?
violet heron
Index i'm guessing is for the script i want the character to say? or can i skip ...
The index is the models accent
You have to upload clean audio for it to convert
frozen harbor
The index is the models accent
Will that file be in the ai folder i downloaded?
torn flare
did the model come with a index in the zip?
Unzipped it
violet heron
Will that file be in the ai folder i downloaded?
Yes
Each model you download contains a .pth and a .index
frozen harbor
Each model you download contains a .pth and a .index
Hmm, let's see if i can find it. What will that be under?
violet heron
Hmm, let's see if i can find it. What will that be under?
The model you downloaded
frozen harbor
The model you downloaded
Like, the entire folder??
violet heron
The goku voice you downloaded was a zip file it should have those inside
The .pth and .index
torn flare
@violet heron can you recommend any vtuber model u know works so I can test to see if its the model or the rvc
torn flare
U know it works ?
violet heron
U know it works ?
Just test a random character from #1175430844685484042
Why does it have to be a vtuber model
violet heron
Do u know any vtuber ones that work
Why does it matter if it’s a vtuber
Just search one you know and test it
proud elbow
Do u know any vtuber ones that work
you'll never know till you have tried 
frozen harbor
The goku voice you downloaded was a zip file it should have those inside
Hey, i think i got it! Thank you so much for your help 👍 💯
torn flare
@violet heron what does the Try loading message mean that it's trying to load index but can't or ?
Does starting http or https make a diffrence as https seems to do more
violet heron
Does starting http or https make a diffrence as https seems to do more
Do start http
Not https
torn flare
Do start http
What's the diffrence ones local and ones server?
rare gobletBOT
Ayo? @torn flare level 2 !!!
violet heron
What's the diffrence ones local and ones server?
Basically
Unless your gonna use it on a another computer or sthm use http
torn flare
Http doesn't seem to do much
So I tried 5 models all either give me the try loading message or generating sola buffer
Gpu I got Is a 1080 founders
@violet heron
violet heron
So I tried 5 models all either give me the try loading message or generating sol...
Are you uploading the .pth inside of the zip or just the zip overall
torn flare
Are you uploading the .pth inside of the zip or just the zip overall
I'm confused
I unzipped it
And uploaded both seperetly
@violet heron I can't just upload the zipitself
torn flare
@violet heron when I use https it gets stuck on sola buffer on http it gets stuck on trying to load index
violet heron
<@858923165379199007> when I use https it gets stuck on sola buffer on http it g...
Don’t load the index then
torn flare
violet heron
Don’t load the index then
People usually have their index at 0
torn flare
What's the index then ?
violet heron
And some models have the accent baked into the pth
torn flare
Make it make sense
violet heron
What's the index then ?
Accent, tho it takes lots of resources and can sound weird
torn flare
Why does it come with index file if it shouldn't be used
violet heron
Why does it come with index file if it shouldn't be used
it’s used in ai covers
torn flare
Now we get index file not found error
violet heron
Do I need to reinstall
Reinstall what
violet heron
Like delete model dir and pretrain
That might fix it
torn flare
I'll try later
rare gobletBOT
Ayo? @torn flare level 3 !!!
cinder ivy
I'd there a way to separate 2 voices from a single audio?
Like if there are 2 people singing
I need them separately
distant turtle
-colab
torn flare
-overtraining
azure marshBOT
-overtraining
You can detect if a model is overtraining if the TensorBoard graph starts to rise and never comes back down. An overtrained model will sound robotic, muffled, and won't be able to articulate words well.
Check these resources to learn more about this topic
- Epochs & TensorBoard from
AI HUB Docs - TensorBoard from 🍏 Applio Docs
plush compass
Does anyone use a pitch algorithm other than rmvpe during learning? Does everyone use rmvpe these days?
grim bison
whenever i try to download my model, after training and all is finished, it keeps giving me an error screen. does anyone know why?
ionic spade
Hello I want to keep the music in the backgound runing , but the RVC just removed it what should I do ?
torn flare
Keep getting index file not found error but some of the models don't come with index
rose pebble
Is there a way to keep training my model after it finished processing the epochs?
Do you mean the D & G files?
rare gobletBOT
Ayo? @rose pebble level 1 !!!
rose pebble
Oh the entire folder
Yes I do have it
I'm still a bit confused, Do I fill out the configurations with the same ones I used in my original and then use load the pretrained base models?
rose pebble
Alright thanks
I js got this error code
ValueError: invalid literal for int() with base 10: 'of'
What does it mean?
gloomy siren
is there a juice wrld rvc
pastel oak
is there a juice wrld rvc
https://weights.gg/models/clm72ktd902c1cctcebf41faa
click on ... and download model
pastel oak
I js got this error code
ValueError: invalid literal for int() with base 10: 'of...
What program are you running, what did you try to do, when does this error come up
this is one of the most confusing errors so you have to try a few things
- rename your folder "Voice Changer" to "VoiceChanger", no spaces inbetween
- keep the original name of MMVCServerSIO instead of "Voice Changer", move the MMVCServerSIO folder inside the VoiceChanger folder if you must
- turn off your antivirus (the normal windows defender is ok to have usually)
- make sure you unzipped the folder (also repeat this step with antivirus turned off, because it can delete files when unzipping)
because you wrote httpS
can you delete the shortcut you made and run the file from inside mmv folder
wispy lodge
This is not the first time I see this and this never ceases to amaze me. Why MODELS and start_http.bat outside the MMVCServerSIO folder?
pastel oak
its from a youtube tutorial
its inefficient and bad
so it is not unzipped
lol
extract it
there is a button that says extract all
else download 7zip
for any other setup stuff you can check this out
rare gobletBOT
Ayo? @brittle wing level 3 !!!
sly atlas
is titian and ov2super works with other language or only english ?
proper shale
is titian and ov2super works with other language or only english ?
All pretrains work on different languages.
sly atlas
All pretrains work on different languages.
thx!
torn flare
what do these do
torn flare
Those are pitch extraction methods
which is best
analog obsidian
which is best
Fcpe is faster than rmvpe tho some models suffer pronunciation issues with it
molten fractal
Hi guys, i know this sounds dumb but how do i download my finished trained model on applio?
torn flare
Fcpe is faster than rmvpe tho some models suffer pronunciation issues with it
so what is best for sounding convincing
analog obsidian
so what is best for sounding convincing
to avoid pronunciation issues just use rmvpe
violet heron
so what is best for sounding convincing
RMVPE
violet heron
Hi guys, i know this sounds dumb but how do i download my finished trained model...
Should be a export pth and index button
I don’t really use applio for training
violet heron
what is onyx then]
For models in .onnx format
Models in onnx sound worse but have AMD support
molten fractal
Should be a export pth and index button
Okay, thanks!
rare gobletBOT
Ayo? @molten fractal level 1 !!!
torn flare
For models in .onnx format
so use rmvpe without _onyx
violet heron
so use rmvpe without _onyx
Yes
torn flare
higher = less noise is converted
n a i want my moans converted
proper shale
n a i want my moans converted
crude flame
n a i want my moans converted
Pardon
scarlet crystal
can anyone please suggest me some good vocal removal app?
i had one but i forgot the name
proper shale
can anyone please suggest me some good vocal removal app?
i had one but i forgot...
MVSep and X-Minus are all decent
That's web tho
There's UVR as well, in case you want to use it directly on your PC
scarlet crystal
MVSep and X-Minus are all decent
well i dont mind using on web but i want to use the best one
i feel like uvr will be better
proper shale
both get the task done, and both have the same models really
pastel oak
n a i want my moans converted
my moans go hard
tough fiber
bro why i cant choose 32k sample 
pastel oak
bro why i cant choose 32k sample <:angry:1169302699699875891>
click v1 then click on v2 again
torn flare
my moans go hard
proof
tough fiber
click v1 then click on v2 again
doesnt work bro i tried 1000 times.
~rvc
~~rvc
-rvc
tough fiber
is there any merge thing except this and okada
azure marshBOT
-realtime
This interaction has expired, use the command
/guides realtimeif you wish to see it again.
warm iris
is there any merge thing except this and okada
like merge voices?
tough fiber
like merge voices?
yea it isnt change only merge thing
warm iris
I use applio, they have a feature for merging.
warm iris
lmme get the compiled link
https://huggingface.co/IAHispano/Applio/tree/main/Compiled
get the 3.2.1 exe
tough fiber
https://huggingface.co/IAHispano/Applio/tree/main/Compiled
get the 3.2.1 exe
trying
pastel lichen
someone help when i put cable as my input audio for discord it doesn't let me talk and im trying to do a voice changer and it only works on discord if i have it as the cable input but it doesn't let me talk.
ashen axle
@earnest musk
violet heron
<@1138318590760718416>
What are you trying to do
bronze maple
Have you found the best way to extract an acapella in the best possible quality?
turbid meteor
weights
pastel oak
Have you found the best way to extract an acapella in the best possible quality?
do you mean splitting acapella from an instrumental?
bronze maple
do you mean splitting acapella from an instrumental?
yess
bronze maple
do you mean splitting acapella from an instrumental?
do you still use UVR, best presets?
pastel oak
do you still use UVR, best presets?
i use mvsep.com and use bs-roformer: 2024.4
bronze maple
i use mvsep.com and use bs-roformer: 2024.4
you have to pay for wav
pastel oak
no
bronze maple
no
thank you I will try
how do you download music in better quality for example spotify, youtube etc.?
pastel oak
how do you download music in better quality for example spotify, youtube etc.?
youtube idk just use anything theres not much
signal summit
can someone help me install rvc?
grim bison
whenever i try to download my model, after training and all is finished, it keeps giving me an error screen. does anyone know why?
glacial pollen
can someone help me install rvc?
Instructions are included on the repo, just please read carefully and you should be good.
signal summit
Try this~
https://github.com/codename0og/codename-rvc-fork
can you come voice with me
to explain pls
glacial pollen
y?
I already told you, have a read.
I wrote those instructions for a reason
AI isn't something you should get into if you can't spare 5-10 minutes figuring out the workflow
glacial pollen
especially if you're being given a semi-automated process of installation that requires you to have a lil read
The info on the repo itself
Just have a read, from the top to the bottom and that's all
rare gobletBOT
Ayo? @glacial pollen level 10 !!!
signal summit
is not working for me when i go to requirements and install my right version when i open the .bat is don't do anything
rare gobletBOT
Ayo? @signal summit level 1 !!!
glacial pollen
that means your python is A) missing or B) installed in a wrong way.
requires 3.10.6
Google stuff up, I only provide an easy way of installing rvc, not really any fundamentals or prerequisites for ML or any other AI frameworks
Other than that, you'll need build tools ( visual studio installer )
and with that few packages:
Mainly, msvc ver 140 and sdk
signal summit
i need to install FFmmpeg too?
signal summit
ok
glacial pollen
python
ffmpeg
build tools - MSVC 140 - vs 2015 ( VER 14.00 ) compilation tools
all other issues or errors you'll encounter ( in console or such ) you can easily google up or search it up in here, through channels. There are or were many solutions to various problems
analog obsidian
all other issues or errors you'll encounter ( in console or such ) you can easil...
btw codename do you know how to fix the fp32 issue in your fork? it only lets you train in fp32 regardless if you enable fp16 in config
glacial pollen
btw codename do you know how to fix the fp32 issue in your fork? it only lets yo...
fp32 is because it's more numerically stable than 16, naturall. single precision vs half precision afteral
my fork's targeting quality over performance / speed
analog obsidian
fp32 is because it's more numerically stable than 16, naturall. single precision...
but training rvc in fp32 didnt caused issues vs fp16?
i remember someone said fp16 is better than fp32 in rvc models
glacial pollen
16 does cause issues but many people either disregard it or ignore it
mode collapses and loss values is one of the primary aspects
but then again, I'm not forcing anyone to use my fork or methodology. Again, it's targeting numerical stability, accuracy and quality over speed
fp16 is using amp, sometimes we get exploded gradients
or other issues
or NaNs too
if you wanna tinker with it, look into config and gpu initialization parts of the code
- that's all i can say, cureenrly busy watching anime
glacial pollen
btw codename do you know how to fix the fp32 issue in your fork? it only lets yo...
tho, wait a minute
you're referring to the newest release from github or previous / test releases?
analog obsidian
tho, wait a minute
you're referring to the newest release from github or previou...
newest
glacial pollen
oh, well, in that case it actually follows the original principle
it is fp16 by default and only applies fp16 -> fp32 logic when necessary
it's in the config.py
whole logic behind it
@analog obsidian Might be you've taken a look at the configs produced during training
Those are to be ignored
Some of the values / data stored in there do not reflect the actual changes or settings
odd shale
Hi cody
analog obsidian
<@775545133448953856> Might be you've taken a look at the configs produced durin...
hmm yeah it says fp16run false when i train a model
their config.json have that regardless if i enable fp16 in inuse config
glacial pollen
their config.json have that regardless if i enable fp16 in inuse config
that's normal
ever since I remember it was like that, unless it was changed in applio or such
glacial pollen
their config.json have that regardless if i enable fp16 in inuse config
aside, you can't change it like that
it is the config.py that controls the events
it checks for memory capacity and generally gpu type
if older / hardware insufficient gpus are detected, it triggers the logic to use fp32
else, it stays fp16 as default
analog obsidian
hmm
analog obsidian
else, it stays fp16 as default
so is possible to edit config.py to always use fp16?
glacial pollen
so is possible to edit config.py to always use fp16?
You get it wrong, it does use fp16 always.
only exception is when and if your hardware just isn't capable of that
fp16, aka, mixed precision in this particular case isn't supported on older older hardware
hence why fp32, more accurate but slower and less memory efficient stability is enforced on weaker hardware
analog obsidian
uhm but im using a 4060
rare gobletBOT
Ayo? @analog obsidian level 26 !!!
analog obsidian
not the best but can handle fp16
obtuse igloo
i followed it and now (new day) i open it again but now its loading super long again, did I acidental redownload stuff now?
brittle mulch
anyone have any tips for a mobile user?
rare gobletBOT
Ayo? @brittle mulch level 1 !!!
low shard
anyone have any tips for a mobile user?
For what ?
brittle wing
can someone help
when i try to train the ai
here
output information is just blank
anyone have a solution for this?
brittle wing
does mangio support arabic?
violet heron
anyone have any tips for a mobile user?
On what?
brave garnetBOT
Made with love by Ilaria 💖
⠀
Settings for Nvidia GPUs 
F0 Det.: rmvpe (suggested for all series)
RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra
Advanced Settings
Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low
⠀
Made with love by Ilaria 💖
brittle wing
what are you using?
RVC
brittle wing
Online? Locally ?
locally
low shard
locally
Be sure you are using one that is in our docs https://docs.aihub.wtf/
Its very weird it asks you for a password
Could you send the model download link?
low shard
https://huggingface.co/Goldeath/MuzanKibutsuji/resolve/main/MuzanKibutsuji32k-TI...
Actually,i just noticed there was a typo in the link, use https://huggingface.co/Goldeath/MuzanKibutsuji/resolve/main/Muzan150e32kTitan.zip?download=true
Where you got that link btw
brittle wing
Actually,i just noticed there was a typo in the link, use https://huggingface.co...
oh
low shard
oh
Prolly the creator renamed the file lol, did you get it from #1175430844685484042 ?
brittle wing
Prolly the creator renamed the file lol, did you get it from <#11754308446854840...
yeds
low shard
yeds
Could you share me the link of the post containing that link? So i can warn the creator about it
brittle wing
Could you share me the link of the post containing that link? So i can warn the ...
i dont have access
to type in there
low shard
i dont have access
Nvm, found it, still thx for telling me tho, will tell him to edit the link so other users wont have the same problem
brittle wing
anyone have a solution for this?
?
brittle wing
Actually,i just noticed there was a typo in the link, use https://huggingface.co...
@frank abyss
low shard
<@159829486987247616>
?
low shard
Ohh, all good bro
patent ridge
what program do I need to train a model?
brave garnetBOT
⠀
Google Colabs 
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
brittle mulch
anyone have any tips for a mobile user?
Uhhh so I separated vocals from instrumental but it still doesn’t sound great. Maybe it’s because I’m using the first site in pins and am not sure completely on what the options do - converting live singing
marble forge
is there any software that you can input lyrics from a song and it will output a subtitle file with the timings of the actual words in the mp3? I am trying to make a lyric video of an ai song i made of me and my friend.
tight lantern
I'm having the same issue as this post in github: https://github.com/w-okada/voice-changer/issues/1182
Which is I can only choose those two options
brittle mulch
Uhhh so I separated vocals from instrumental but it still doesn’t sound great. M...
Like in the help document it asks me to add reverb but the site in pins doesn’t allow me to do so since the options are pretty limited (maybe it’s not beginner friendly)
I also think there may be more that I’m missing even from that since there are a lot of things I’m unsure about as to how much work AI covers rlly take
boreal steeple
When I run a crepe extract with Applio it uses my entire 32 gb of memory. Is there a way to avoid this memory leak?
nocturne quest
What’s up with the ai? Every-time I go to record it always crashes?
analog obsidian
When I run a crepe extract with Applio it uses my entire 32 gb of memory. Is the...
reduce cpu cores to 2
spiral osprey
Can someone send me the link to the vocal isolation docs there.
torn flare
https://youtu.be/vBomoqNFUns how do u get a voice this passable
marble forge
is there any software that you can input lyrics from a song and it will output a...
Hello someone please answer
fossil crag
does anyone know how to use the voice models in something like discord to show friends?
vale shadow
I found some ai voices but the download links don't work. WHat do I do
gusty kestrelBOT
@vale shadow, I have found 13 results that match your search!
Requested by warrendemartini
ionic spade
Can I run the RVC just on cpu?
astral jungle
What does the ValueError: not enough values to unpack (expected 2, got 1) error mean?
zealous shuttle
any uvr5 colab that can separate some songs only? (not in folder)
jaunty kayak
Is there a way to run RVC or Applio locally without GPU only using the CPU on a low-resource laptop without having errors or a correct way to do it?
When I tried it, everything was going fine, the peoblem is when it started training the epochs, the screen went black for a few seconds and when the image came back it showed that it had been disconnected.
low shard
what program do I need to train a model?
RVC, check https://docs.aihub.wtf/
low shard
is there any software that you can input lyrics from a song and it will output a...
Well, that outputs a subtitle file not really sure, but speaker diarization could help you maybe https://huggingface.co/spaces/Delik/pyannote-speaker-diarization-3.1 where it will tell the time stamps of the voice in the audio file
low shard
Uhhh so I separated vocals from instrumental but it still doesn’t sound great. M...
What site are you using
proud elbow
Is there a way to run RVC or Applio locally without GPU only using the CPU on a ...
!howtoask
patent trellis
proud elbow
as I said before, explain the issue/errors
low shard
does anyone know how to use the voice models in something like discord to show f...
Yes, its realtime voice changing, so using the models on realtime in calls, u can use things like wokada
-rt
azure marshBOT
-rt
This interaction has expired, use the command
/guides realtimeif you wish to see it again.
low shard
I found some ai voices but the download links don't work. WHat do I do
Share what are you using, and also the voice models download link
low shard
Can I run the RVC just on cpu?
Technically you can inference locally on cpu, but its slow, and u cant train, its better if you just use cloud free gpu like google colab tbh
ionic spade
Technically you can inference locally on cpu, but its slow, and u cant train, it...
I just want to test it
is there any guid or something to help me do it?
low shard
is there any guid or something to help me do it?
Uhh not really sure as i don't do it locally but you could check https://docs.aihub.wtf/ for the local rvcs
ionic spade
Uhh not really sure as i don't do it locally but you could check https://docs.ai...
ty
low shard
any uvr5 colab that can separate some songs only? (not in folder)
Ye there is an uvr colab https://docs.aihub.wtf/rvc/resources/vocal-isolation/#cloud-uvr
zealous shuttle
uhh i wonder, can we move/copy file in cell on google colab?
glad zealot
uhh i wonder, can we move/copy file in cell on google colab?
wot
you can move stuff using the file manager on the right side on colab
vale shadow
Share what are you using, and also the voice models download link
I used the search function and found this model. The link takes me to applio and there if I click download, it says entry not found....
https://applio.org/models/download/1192788502946005013
low shard
I used the search function and found this model. The link takes me to applio and...
The model has been deleted or privated, you can't do anything about it
vale shadow
The model has been deleted or privated, you can't do anything about it
Ah ok
Any other sites where I could search?
low shard
Any other sites where I could search?
The problem wasn't the site applio, its the creator of that specific model that deleted/privated it
Btw
You can search rvc ai voice model:
- #1175430844685484042
- #🔍│find-models
- https://weights.gg/
- https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
- https://applio.org/models
- https://voice-models.com/
if there isnt one, you can:
- #1159289738314919936
- #1191429836321849435
- make it yourself with our docs guides https://docs.aihub.wtf/
vale shadow
The problem wasn't the site applio, its the creator of that specific model that ...
Yeah I meant if I could find other versions created by other people. Thanks though
low shard
Yeah I meant if I could find other versions created by other people. Thanks thou...
Yea i shared to you other sites for searching models
vale shadow
Yea i shared to you other sites for searching models
Yeah thanks again
mystic marsh
-colab
low shard
Yeah thanks again
Yw
brittle wing
does anyone help with applio?
happens with inference and tts
it just creates an empty audio file
proper shale
anything in cmd/command prompt when you convert?
if only i knew how to deal with Applio errors 💀
brittle wing
anything in cmd/command prompt when you convert? <:nails:1159569314848972891>
wait ill check
proper shale
brittle wing
lmao
"invalid value encountered in divide"
does it have to do something with it
?
proper shale
could this be a folder placement issue perhaps? i see it's inside the user folder, maybe if you put it on the downloads folder or something similar
if this is it I am gonna cry 😭
proper shale
I haven't seen this b4
brittle wing
ill try to figure it out ig
if i dont have the gradio thing installed does it affect it?
proper shale
not really. applio normally comes prepackaged with everything unless you installed the non precompiled version ofc
rare gobletBOT
Ayo? @brittle wing level 4 !!!
brittle wing
but i moved it to downloads
hushed sedge
hi
it took my computer like 15 minutes to make a model with 500 epoches is this normal or is there something weid wierd going on with the model or somethig?
because when i trained models last year it took some hours to train 500 epoches
rare gobletBOT
Ayo? @hushed sedge level 4 !!!
pastel oak
it took my computer like 15 minutes to make a model with 500 epoches is this nor...
how long is the dataset
and what pretrain
and what gpu
hushed sedge
how long is the dataset
around 16 minutes
hushed sedge
and what gpu
4070 ti
hushed sedge
and what pretrain
sorry but whats a pretrain
pastel oak
xd so youre using original pretrain
ehmm well thats still kinda fishy
check the graphs if theyre alright, check how the model sounds
if nothings weird youre fine
hushed sedge
i already closed the graphs but it got to like 2000 or maybe 3000 steps i think in this time
and to a smooth of 37-36 i think
which looks normal
its only the time which makes it sus
proud elbow
it took my computer like 15 minutes to make a model with 500 epoches is this nor...
that means average 1.8 seconds per epoch, including checkpoint saving. have you found something wrong in the console log, tensorboard, as well as the model testing?
hushed sedge
that means average 1.8 seconds per epoch, including checkpoint saving. have you ...
i already closed it so i cant check
proud elbow
i already closed it so i cant check
nah you can still check the tensorboard and also test the model, unless you lost them all
hushed sedge
nah you can still check the tensorboard and also test the model, unless you lost...
how
hushed sedge
how are the model weights? you can use them in the inference
yes i can use the model
proud elbow
yes i can use the model
does it work? (I mean the output)
hushed sedge
yes the output works and it doesnt sound distorted
what batch size is it?
proper shale
probably fine lol
hushed sedge
<:PES3_SusAlt:523045516452298762> what batch size is it?
4
teal dome
is something wrong with applio? uploading pth/index just stalls at 0%
it doesn’t even seem to do anything on the colab page
oh… nvm it uploaded. the loading icon indicator just doesn’t move anymore
void pond
A friend has been getting a weird "utf-8 parse" error while trying to train a model
His user path has an acute accent (Usuário), could that be the "utf-8" parse error he has been getting?
proper shale
yep, most likely
odd shale
A friend has been getting a weird "utf-8 parse" error while trying to train a mo...
That's the reason probably.
void pond
Its weird it didn't happen when pre-training, getting feature and index training, only when training the model it gets the error
he's sending everything to C root to see if it fixes, I'll report soon
odd shale
yep, most likely
ayuwoki
brave swallow
I have big issue :
with Xtts each time I generate text in arabic the output WAV has missing parts
like skipping entire sentence
robust meadow
Should I low-cut the dataset?
brave swallow
what you mean
robust meadow
When creating a data set, should I use an equalizer to clean up the very low frequencies?
proud elbow
When creating a data set, should I use an equalizer to clean up the very low fre...
do you mean the high pass method to remove/reduce DC offset?
zealous shuttle
you can move stuff using the file manager on the right side on colab
you mean left side?
zealous shuttle
i can't download it, help
brittle mulch
What site are you using
I was using the first one, Ilaria in pins
rancid sluice
i need help
@proper shale
mano
eu tava falando com um helper gringo mas ele fugiu kkkkk
proper shale
tu nem precisa disso pra rodar o rvc realtime já vem tudo pre-instalado
rancid sluice
did i get pinged
sorry, It was me, I thought the other helper wouldn't come
hushed sedge
whats better to use contentvec or hubert?
analog obsidian
whats better to use contentvec or hubert?
rvc uses contentvec, for some reason rvc-boss renamed it "hubert" tho is not actually hubert
tldr; use contentvec
rancid sluice
@viscid moss help
rancid sluice
rvc uses contentvec, for some *reason* rvc-boss renamed it "hubert" tho is not a...
help
rancid sluice
ok
analog obsidian
rename the folder to something like mymodels
and also remove spaces from the model name
rancid sluice
and also remove spaces from the model name
thxx
rancid sluice
and also remove spaces from the model name
It's very noisy, my voice barely comes out
analog obsidian
It's very noisy, my voice barely comes out
you probably have a lot of background noise
realtime_gui can't do proper noise removal
you either use the one from okada or download steelseries
also is your gpu nvidia or amd?
rancid sluice
you probably have a lot of background noise
No, my microphone is good
my voice is lagged
i dont speak english well
analog obsidian
my voice is lagged
sample length is too low, increase it to 0.30
rancid sluice
also is your gpu nvidia or amd?
amd
analog obsidian
amd
oh that explains
amd works poorly in realtime_gui
use this for amd realtime voice changing
rancid sluice
ok
analog obsidian
for amd download voice-changer-windows-amd64-dml.zip
rancid sluice
right
hushed sedge
Hiii question
My computer takes around 18 seconds and 100 steps to make an epoch is that a normal rate?
proper shale
My computer takes around 18 seconds and 100 steps to make an epoch is that a nor...
Seems normal
analog obsidian
My computer takes around 18 seconds and 100 steps to make an epoch is that a nor...
18 seconds per epoch? yes seems normal
brittle mulch
What site are you using
oh yeah it may also be because of the voice models, the ones I have found for MLP have had 0 likes so far so maybe that’s also part of the priblem
hushed sedge
Seems normal
Ok thanks 
proper shale
Ok thanks <:TextW:1159654826771230741>
You're welcome!
rancid sluice
Seems normal
o outro helper falou q o rvc e ruim pro meu pc mano
rancid sluice
for amd download voice-changer-windows-amd64-dml.zip
ele falou pra baixar esse
rancid sluice
o outro helper falou q o rvc e ruim pro meu pc mano
falei errado, mas acho q vc entendeu
e esse q ele falou pra mim baixar n funciona kkkkkk
rare gobletBOT
Ayo? @rancid sluice level 4 !!!
proper shale
😭
proper shale
tu tem q selecionar o modelo
rancid sluice
mds
proper shale
e muda o f0 pro crepe_tiny
proper shale
de nada
proper shale
honestamente n sei
proper shale
sim, mas tem alguns problemas que eu não sei resolver
rancid sluice
oq vc faria?
wispy lodge
Check command line that the app opens. If there are any issues, it should show errors
If there are no errors, maybe you selected wrong microphone. Or your microphone is muted
proper shale
ah
wispy lodge
Could you scroll up? There as an error above this message
rancid sluice
wispy lodge
Restart the app
wispy lodge
i think deleting pretrain and model_dir might work tho
This should never be the case in my version. Files are validated before start and it won't launch if files are broken
rancid sluice
low shard
I was using the first one, Ilaria in pins
the hugging face space ilaria rvc ?
stark wadi
I'm trying to download RVC on mac but I keep getting an aria2 error when trying to use run.sh Any pointers?
wispy lodge
Which model did you use? Looks like something is wrong with it
low shard
oh yeah it may also be because of the voice models, the ones I have found for ML...
yea could depend by the models, be sure that the vocals are very clean and that play around with the pitch settings
wispy lodge
No, that's app. I mean, voice model like #1175430844685484042
rancid sluice
faustao
proper shale
This should never be the case in my version. Files are validated before start an...
Ah
analog obsidian
if your model sound bad no matter what then is the model problem
rancid sluice
okay
wispy lodge
Or try different voice model
rancid sluice
rancid sluice
Or try different voice model
ok
wispy lodge
https://applio.org/models/1145363219448336424
Looks like a broken voice model (or maybe RVC v1 model?). Doesn't load for me too
Yeah, that's RVC v1 model. Those don't work in my version :\
rancid sluice
Yeah, that's RVC v1 model. Those don't work in my version :\
hey
wispy lodge
this error remains
Delete stored_setting.json and restart the app
And don't select this model
rancid sluice
ok
brittle mulch
the hugging face space ilaria rvc ?
ye that one, not sure if I should use another one if I’m just starting out bc of the settings
brittle wing
https://imgur.com/a/JrDU5iN Does anyone know why I can only select the two and my graphics card is not on the right but only cpu gpu0 gpu1 and so on
rare gobletBOT
Ayo? @brittle wing level 1 !!!
proper shale
https://imgur.com/a/JrDU5iN Does anyone know why I can only select the two and m...
That's basically the quirks of the AMD version
At least the original one in that case
rancid sluice
And don't select this model
work
brittle wing
in the video the guy say for nvidia
rancid sluice
but, the voice is so lagged
brave garnetBOT
⠀
Download for Nvidia GPUs
Version 18a cuda
Download for AMD GPUs 
Version 18a directml
Download for Intel GPUs 
Version 18a directml
Download for Mac 
Version 17b Mac
⠀
Made with love by Ilaria 💖
proper shale
@brittle wing download the NVIDIA version there
brittle wing
send me pls is all the same
wispy lodge
but, the voice is so lagged
Could you show your settings and these numbers when you speak?
proper shale
send me pls is all the same
wispy lodge
- Set Extra to 1s (more extra, better quality).
- Set Gain IN to 1.5.
- Set S. Thresh. to 0.00015
- Change Tune. This is a pitch of your voice.
- Set monitor to none (you hear yourself twice because of this)
Also, click Stop before you change settings
rare gobletBOT
Ayo? @wispy lodge level 27 !!!
wispy lodge
Then Start when done changing settings
rancid sluice
ok
low shard
ye that one, not sure if I should use another one if I’m just starting out bc of...
Nono, its all good, its the faster cloud inference space so, you might need to use another model, or play with the settings till it sounds good (the settings are explained here )
brittle wing
https://huggingface.co/wok000/vcclient000/blob/main/MMVCServerSIO_win_onnxgpu-cu...
ist for nvidia i have gtx 1070
@proper shale
proper shale
yes, the version i sent you is for Nvidia
gusty kestrelBOT
@rancid sluice, I have found 5 results that match your search!
Requested by zipzapo
rancid sluice
1. Set Extra to 1s (more extra, better quality).
2. Set Gain IN to 1.5.
3. Set S...
It’s improved, is there a good female model out there for me to test? The ones I'm testing look like a monstrosity, I don't know if it's lag
open musk
hi wasn't here for a long time so uh what's the site to load pth and index files?
rare gobletBOT
Ayo? @open musk level 1 !!!
wispy lodge
It’s improved, is there a good female model out there for me to test? The ones I...
For female models, you should use 9-12 tune. Depends on how low your voice is.
This one sounds ok
https://discord.com/channels/1159260121998827560/1229978221655756911
lone grove
I installed rvc and it doesnt apply on discord
is it not real time???
sharp field
does any1 have a good vocal remover? UVR is kinda ass ngl
rancid sluice
For female models, you should use 9-12 tune. Depends on how low your voice is.
T...
wow
rancid sluice
For female models, you should use 9-12 tune. Depends on how low your voice is.
T...
Thank you very much bro, you showed a lot of patience from everyone here, I really appreciate your attention
analog obsidian
does any1 have a good vocal remover? UVR is kinda ass ngl
mvsep
rancid sluice
For female models, you should use 9-12 tune. Depends on how low your voice is.
T...
how can i put my voice in discord?
sharp field
oh pog
rare gobletBOT
Ayo? @sharp field level 2 !!!
analog obsidian
analog obsidian
mystic marsh
-colab
proper shale
<:sitt:1118547278861443182>
have you created an account
i definitely recommend doing it, since it reduces queues massively
sharp field
oh ok
proper shale
unless ofc the servers r clogged
sharp field
proper shale
😭
sharp field
proper shale
that's better
proper shale
yippee
sharp field
is charlie puth's we don't talk anymore just a hard song to cover?
because istg im dying
gusty kestrelBOT
@rancid sluice, I have found 1 results that match your search!
Requested by zipzapo
feral saffron
-guide
gusty kestrelBOT
@rancid sluice, I have found 1 results that match your search!
Requested by zipzapo
wispy lodge
how can i put my voice in discord?
You need to install VAC.
https://rentry.co/VoiceChangerGuide#virtual-audio-cables-mandatory
Then you set Audio output to Line 1 (Virtual Audio Cable) in voice changer. In Discord, you set Audio input to Line 1 (Virtual Audio Cable)
gusty kestrelBOT
@rancid sluice, I have found 1 results that match your search!
Requested by zipzapo
@rancid sluice, I have found 1 results that match your search!
Requested by zipzapo
kind meadow
how do i get it t sound realistic?
proper shale
I install for nvidia but its AMD?
dw about the "AMD64" stuff, the GUI is different so it's the NVIDIA version
proper shale
how do i get it t sound realistic?
mess with tune, index & extra...
but eh it might not sound convincing anyway, since it's AI
proper shale
how to use ilaria rvc
https://docs.google.com/document/d/1YbXcLFPaGjhOdG5NFkK3QrucCEpHZBwFUxkeMO8aB18/edit
if you mean the huggingface version
outer cosmos
thank youuu
mystic pawn
-colab
brittle basalt
-colab
undone steppe
hi, there is a new gui software?
the rvc-gui has a lot of bugs and very slow, at least for me
kind meadow
wh y cant i put crepe only crepe tiny
rare gobletBOT
Ayo? @kind meadow level 1 !!!
gusty kestrelBOT
@rancid sluice, I have found 1 results that match your search!
Requested by zipzapo
@rancid sluice, I have found 2 results that match your search!
Requested by zipzapo
zealous shuttle
anyone had a apple music account that not used? i want to download a song from it using downloader
rancid sluice
oh, sorry
rare gobletBOT
Ayo? @rancid sluice level 5 !!!
brittle wing
Any one knows about a good program for RVC for Mac m1 can’t find one that works properly with terminal even with python and homebrew installed and xcode
brave garnetBOT
Made with love by Ilaria 💖
brittle wing
Yeah its work perfekt thanks perfect*
spiral osprey
Why don't I import my audio to the website?
zealous shuttle
rare gobletBOT
Ayo? @zealous shuttle level 13 !!!
zealous shuttle
does doubledouble didn't work anymore?
lucid solar
@potent badger 👋 #1245889671515344987 message
Я установил твою модель, помоги правильно настроить🙏
patent ridge
is there a website that trains a model from dataset?
I may or may not be stupid but downloading a bunch of crab and/or running google collab is too complex for me
thin stump
is there a website that trains a model from dataset?
You can try weights.gg. The quality would be the same here but it all depends on your dataset
ok, fair
violet heron
should i download the directml?
Yes
violet heron
wh y cant i put crepe only crepe tiny
GPU?
violet heron
what is the index supposed to do ?
Acts like accent, a lot of people don’t use it in realtime since it can make the voice more laggy or take up more performance
muted cosmos
thank you
rare gobletBOT
Ayo? @muted cosmos level 1 !!!
muted cosmos
and one more thing
is it normal for my cpu to go 80% and gpu only 14% ?
even tho i selected gpu 0
my gpu
pallid rampart
Help me please! I open the start_http.bat file to install the application, but the command prompt opens for 2 seconds and disappears. can't download :_(
narrow bobcat
Is there a DML version for the non realtime rvc?
potent badger
<@388018204431351809> 👋 https://discord.com/channels/1159260121998827560/12458...
в рус чате есть гайд по настройке самого вч, а под мою модель лишь tune крутить надо, тут зависит от твоего голоса
pallid rampart
в рус чате есть гайд по настройке самого вч, а под мою модель лишь tune крутить ...
помоги мне пжж
quaint nacelle
Looking for someone who loves singing lil peep and can articulate well like him . Will pay money! Doing a secret project
lucid solar
в рус чате есть гайд по настройке самого вч, а под мою модель лишь tune крутить ...
когда можешь зайти в гс
Мы тупим
pallid rampart
в рус чате есть гайд по настройке самого вч, а под мою модель лишь tune крутить ...
и мне помоги пжж
zenith raven
somebody can help with train model
scarlet venture
#buscador
rare gobletBOT
Ayo? @scarlet venture level 1 !!!
brave garnetBOT
Made with love by Ilaria 💖
glacial pollen
Is there a DML version for the non realtime rvc?
Unfortunately, kinda no
Unless it was reverted and dml "support" is back to mainline or any other forks, it's faulty
causes pitch drift and other issues + memory issues
You have to do your research, get into linux and use rvc with rocm to utilize amd, if that's your goal
if your goal, on the other hand, is to use dml on cpu? yeah no
glacial pollen
is it normal for my cpu to go 80% and gpu only 14% ?
Truth to be told, gpu utilization isn't high for that, it is the memory consumption and allocation
cpu naturally will be doing the job under the hood ( even more if you use harvest which is pure cpu or older f0 methods )
also, index [ faiss ] ( in case you use it ) is fully cpu dependent
- pm, dio and harvest: they use cpu
- crepe, rmvpe, fcpe: use gpu ( not 100% sure but might be there's cpu variant of fcpe? don't remember. )
- index / faiss: uses cpu heavily
Easiest way to debug whether your gpu's utilized properly or not is check for memory usage / allocation for your gpu.
wispy lodge
causes pitch drift and other issues + memory issues
It does with long audio chunks (and rvc gui splits into 32s chunks by default iirc), but it maintains more or less correct pitch on short chunks (less than 5s). Which is how it works in realtime at least
stark wadi
Whenever I try to train a model with applio, I get the following error in the terminal. Traceback (most recent call last) :
File "C: . py",
hps = get_hparams()
File "C: l\rvc\train\utils .py",
with open(config_save_path, as f:
line 54, in <module>
line 258, in get_hparams
FileNotFoundError: [Errno 2] No such file or directory: 'C:
Failed to train index: [WinError 3] The system cannot find the path specified: 'C:
\ logs \ \my- project \ \ 3_feature768 '
oh. I had it in downloads folder under my user.
I tried to put it in the C:\ directory, but I still get the same error when starting training. The dataset is in the dataset folder, under a folder called test. Within the test folder are all of the audio files.
so the full path to the data set is: "C:\applio\Applio\assets\datasets\voice"
but I still get the error.
proper shale
so the full path to the data set is: "C:\applio\Applio\assets\datasets\voice"
aren't you on Mac
or did you switch
stark wadi
I am using a windows PC to train.
proper shale
ah i see
stark wadi
I have both mac and windows.
proper shale
nice
proper shale
Whenever I try to train a model with applio, I get the following error in the te...
honestly, I'm not sure what's up... the path in the log is all screwed up as well... just weird
stark wadi
do you mind if I hop on a call with you? That is what I could grab from the OCRing of the terminal window, as my screen reader didn't want to read the terminal output, for some reason.
proper shale
do you mind if I hop on a call with you? That is what I could grab from the OCRi...
can't go on calls sadly...
stark wadi
so it might've been messed up, I mean, the text that I copied.
proper shale
so it might've been messed up, I mean, the text that I copied.
maybe it's that yeah
honestly... I'd probably use RVC Mainline instead for that
stark wadi
I tried that as well. Got a similar error.
With RVC mainline, it trained once and then it wouldn't train again for some reason. I thought the RVC mainline wasn't being updated anymore?
proper shale
i mean mainline still works, but yeah it hasn't had a big update for months at this point
idk what's up really
glacial pollen
It does with long audio chunks (and rvc gui splits into 32s chunks by default ii...
we ain't talking about gui
but rvc itself with dml
gui shouldn't be taken into account in this situation
proper shale
idk if it's a problem related to placing it in the root of C: drive...
stark wadi
I am currently re-downloading rvc mainline. Hopefully this works. Will update once downloaded and trying to train.
wispy lodge
but rvc itself with dml
Well, ok. DML does not handle long audio well with RVC
Because some calculations drift/are unstable on that backend for some reason
urban wasp
-colab
stark wadi
got it training with mainline.
proper shale
Oh wow
brave garnetBOT
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
onyx isle
I am having the same problem as this guy: https://www.reddit.com/r/apps/comments/17okjjw/rvc_gui_traceback_most_recent_calls_error/
proper shale
don't use RVC GUI - it's outdated.
prefer newer local options like Mainline and Applio
https://docs.aihub.wtf/ - RVC > Local section
onyx isle
Thanks
proper shale
You're welcome, let me know how it goes :)
dim anchor
https://imgur.com/a/Vf8aLgC hello can anyone help me? i dont know what is this
azure marshBOT
Suggestions for @dim anchor
zealous shuttle
does uvr colab support audio with m4a format?
abstract kettle
What's epochs
dim anchor
thx
zealous shuttle
how to cut the audio from files in colab?
i mean moving the audio to another folder after separating the vocals
violet heron
What's epochs
Training cycles, more or less doesn’t mean better
zealous shuttle
no no....i mean when after separating the vocals, i need to separate the backing vocals
violet heron
Training cycles, more or less doesn’t mean better
You can ignore if your looking for a model
zealous shuttle
which is the colab can't select a specific file
abstract kettle
Training cycles, more or less doesn’t mean better
How i can do this
violet heron
How i can do this
You don’t, that’s just information about the model, ignore it if your looking for models
zealous shuttle
ohh so....i just do a separate once only?
zealous shuttle
no no....i mean when after separating the vocals, i need to separate the backing...
separate the backing vocals, remove the echo effect
Ahh i see......guess i just gonna looking other colab
zealous shuttle
One by one still fine by me
Separating voc - Inst --> Separating voc - backing --> Remove Echo fx
Wish someone made uvr colab that can select a specific audio and had a complete separation
mystic pawn
-colab
narrow bobcat
Unfortunately, kinda no
Nevermind, I found a work around to work on dml just for inference purpose and got it to work.
rare gobletBOT
Ayo? @narrow bobcat level 1 !!!
glacial pollen
Nevermind, I found a work around to work on dml just for inference purpose and g...
well good luck, it's still broken nevertheless
mystic pawn
yo anyone has a good colab for inference¿
rare gobletBOT
Ayo? @mystic pawn level 4 !!!
mystic pawn
xd
fleet trail
lmao
mystic pawn
How I make a ai voice my that "RVC" thing
It's a whole process, you need to have 4-10 minutes in voice clips for a good ai model
then you need to train all that, blah blah blah and that's it
fleet trail
Damn
mystic pawn
it's not hard but it's tedious
fleet trail
I just wanted to do a kpop cover🙏
mystic pawn
I just wanted to do a kpop cover🙏
Oh, you can search if that singer already has its own existing model in #🔍│find-models
or in the voice model tab you can search in the search bar
fleet trail
I already have the singer's link
mystic pawn
Are you on PC?
fleet trail
There is a part that gives an error
fleet trail
Are you on PC?
Phone
fleet trail
It's ok
low shard
Outdated colab
Do you want to train model (make ai voice model) or to inference (use an ai voice model)
fleet trail
It's like expired?
low shard
It's like expired?
Its outdated, that colab you used is old and not updated anymore
low shard
Do you want to train model (make ai voice model) or to inference (use an ai voic...
i can give you a more updated colab if you can answer me this
fleet trail
Do you want to train model (make ai voice model) or to inference (use an ai voic...
Inference
proven hill
Want to make an AI Cover?
You might wanna use Ilaria RVC Zero! 🎉
It runs on Hugging Face’sZeroGPU, which currently, is an A100 GPU (11 times faster than Google Colab’s T4 Free daily GPU),
with 69 GB of RAM, the fastest free cloud RVC!If it's helpful, put a like to the space to help Ilaria!
Link: https://huggingface.co/spaces/TheStinger/Ilaria_RVC
Guide: https://docs.google.com/document/d/1YbXcLFPaGjhOdG5NFkK3QrucCEpHZBwFUxkeMO8aB18
low shard
# Want to make an AI Cover?
### You might wanna use Ilaria RVC Zero! 🎉
> It ...
yea ilaria answered pretty much @fleet trail
fleet trail
ilaria my savior
proven hill
low shard
Ilaria RVC on Colab is NOT WORKING, BUT there's a BETTER & NEW ONE!
What do I use instead?
- For VERY FAST INFERENCE: Ilaria RVC Zero + The Guide - The fastest free RVC Hugging Face Space for Inference (Making your ai cover!) ,It runs on Hugging Face’s ZeroGPU, which currently, is an A100 GPU (11 times faster than Google Colab’s T4 Free daily GPU), with 69 GB of RAM
Unfortunately you can't train with ilaria rvc on cloud
as the hugging face space got fixed and improved (ilaria rvc zero), ima unpin this and pin a newly message with it too (sorry, just doing it bc people come still asking for help with outdated colabs lol)
brittle wing
i used ultimate vocal remover on a 2 hour long file and its nearly at the end but my whole pc is frozen
should i restart or leave it
there is alot of background music thru most of it
if its still messed up in like 10 mins il just cut the parts w music and do them seperatelg
i have a good gpu so its better if i do it on my@pc i think
Yeah its lowkey just cuz im too lazy to cut it into parts
wait my pc unfroze
yeah
i didnt know u could do multiple files at once until you literally just mentioned it right now
thats why i didnt do it
yes
Will it remove like sound effects like clicks and stuff or nah
nvm its saving the stems rn il find out for myself 
alot of it is just silence
lol
once i do all the labels and clean it up it will prob be 20-40 mins
oh my god i forgot to select vocals only its gonna take like a year to save the instrumental
i already waited it out theres no point of cancelling now
nvm it was pretty quick
time elapsed: a hour and 30 mins
it actually worked really well tbh most the unneccessary noises r gone
uncut surge
cursive elk
Okay so I am having a very hard time getting started and I cant find a good guide that helps because I am small brain
rare gobletBOT
Ayo? @cursive elk level 1 !!!
abstract locust
hello I am set to rvc but when I use one see it is robotic giga its normal
sharp field
What's an index file?
kind meadow
why can i only pick these too
analog obsidian
why can i only pick these too
amd can only choose onnx for f0 extractor
kind meadow
amd can only choose onnx for f0 extractor
so i can only pick those too?
analog obsidian
so i can only pick those too?
yes, anyways, is that official wokada?
for amd users modded wokada is the best
and also has onnx versions of crepe and fcpe
but besides that, is just faster in the modded version
lusty kindle
cutting out a lot even with it on 192 chunks (never did this before)
16384 for extra
its fine on 256 chunks tho
kind meadow
but besides that, is just faster in the modded version
wait so whats better onnxgpu or onnxdirectxl
analog obsidian
wait so whats better onnxgpu or onnxdirectxl
directml for amd