ok no problem
#✨│ai-help
frosty kayak
spare shale
But I think you can say this to those who made the site? 😅
brittle wing
I've forwarded it just ye
fickle delta
all good lol, didn't mean to make it sound harsh or anything
spare shale
I've forwarded it just ye
?
brittle wing
All good, thank you :)
fickle delta
awesome!
spare shale
I have forwarded it to website people.
Thank you and forgive me 😅
spare shale
gonna go back to crying now
spare shale
gonna go back to crying now
Don't be sad and don't cry ❤️
Hold the cookies 🍪
frosty kayak
hey charlotte, have you ever continued training after collapses and gotten a good model? its set at batch size 6 rn but i already tried batch 8 and was having some too
lusty void
do i turn on passthrough for voice changer
i hear myself but my voice isnt changing
molten pecan
That's the old dead website
https://aihubdocs.github.io/en/ this is the new one
It's in early development
frosty kayak
new website is cool, easy to maneuver
vast forge
what if my dataset is 4 minutes long?
frosty kayak
if you can shoot for 15 mins - 30 mins
vast forge
what happens if i copy and paste
half cove
If you duplicate the same data it will have the same result
half cove
what if my dataset is 4 minutes long?
Try using ov2 and lower batch size
vast forge
ok
half cove
Get his permission 
vast forge
the audio was 19 minutes long until i removed the silence and it brought down to 4 minutes
vast forge
Get his permission <:nails:1159569314848972891>
i did
snow veldt
anyone know how to fix this D:\Program Files\Applio\Applio-3.0.4\env\lib\site-packages\torch\autograd\__init__.py:251: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance. grad.sizes() = [64, 1, 4], strides() = [4, 1, 1] bucket_view.sizes() = [64, 1, 4], strides() = [4, 4, 1] (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\torch\csrc\distributed\c10d\reducer.cpp:334.) Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
old oracle
what notebook do I use for traning?
vast forge
why dont i have a index file
full tundra
how to make ai cover with only vocal audio, but google collab considers part of the file to be instrumental and not vocals?Simply put, I want to turn off the separation of vocals and instrumentals(i use AICoverGen WebUI (Modded by Hina))
azure marshBOT
how to make ai cover with only vocal audio, but google collab considers part of ...
forest sluice
how to make ai cover with only vocal audio, but google collab considers part of ...
Maybe try out Ilaria RVC
forest sluice
-colab
forest sluice
Or this Colab: https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing
full tundra
thx
rare gobletBOT
Ayo? @full tundra level 1 !!! 
eager condor
thanks
forest sluice
My improved version of the dl_models.sh (download models script, for Mac/Linux), which uses curl (pretty much commonplace for UNIX-like systems): https://gist.githubusercontent.com/sonphantrung/66c5eed963296ca519bc89872e3788da/raw/15f40fb326b46ca4b633554a5d18df9c0e6cb170/dlmodels_improved.sh
How to use:
- Right click, save the file to the RVC folder
- Run
chmod +x ./dlmodels_improved.sh - Finally, run
./dlmodels_improved.sh
forest sluice
My improved version of the `dl_models.sh` (download models script, for Mac/Linux...
I'm sorry that I forgot to add the -L option in the curl command 😭
random ice
which path file do i grab for the models
frosty kayak
i usually make sure all 4 is in the zip....G, D, Model name file w/ last epoch, and index/added file
silent sun
Do you need any software programs to download the app? Like
Java or python
random ice
i usually make sure all 4 is in the zip....G, D, Model name file w/ last epoch, ...
i only have the g and d im pretty sure
rare gobletBOT
Ayo? @random ice level 1 !!!
silent sun
Why can't I open the App on Mac Os
rare gobletBOT
Ayo? @silent sun level 1 !!!
frosty kayak
i only have the g and d im pretty sure
I use rvc disconnected but if there is no model file....definitely have the G, D, and added....that is a MUST
random ice
I use rvc disconnected but if there is no model file....definitely have the G, D...
how do you get a model file?
silent sun
Did I download the wrong app? MMVCServerSIO? for Mac OS?
Can someone help me for once?
frosty kayak
how do you get a model file?
run the cell that says, "Export model from notebook to drive"
frosty kayak
Can someone help me for once?
i use colab online, sorry man
random ice
run the cell that says, "Export model from notebook to drive"
is that the export Onnx
frosty kayak
is that the export Onnx
if you are using rvc disconnected it is the next step after training
random ice
what which is rvc disconnected
frosty kayak
-colab
random ice
yeah i dont use the disconnected one
frosty kayak
yeah i dont use the disconnected one
first group, second to last
frosty kayak
yeah i dont use the disconnected one
just type "-colab"
random ice
can i get an image or something i am very lost here
frosty kayak
can i get an image or something i am very lost here
random ice
im talking about getting the model.pth file from my existing model not in the disconnected
or do i have to use the disconnected to get the model.pth
stiff helm
I'm sorry that I forgot to add the `-L` option in the curl command 😭
That’s alright, I think it downloaded correctly now
@forest sluice I’m still having issues now though, I’m not sure if you’ve seen above
silent sun
do i need python to run the application on mac?
oak barn
I need help. I made the model but I am having trouble with the cover. I get this error:
~] Converting voice using RVC...
gin_channels: 256 self.spk_embed_dim: 109
<All keys matched successfully>
Traceback (most recent call last):
File "/content/FIX/src/main.py", line 291, in song_cover_pipeline
voice_change(voice_model, main_vocals_dereverb_path, ai_vocals_path, pitch_change, f0_method, index_rate, filter_radius, rms_mix_rate, protect, crepe_hop_length, is_webui)
File "/content/FIX/src/main.py", line 201, in voice_change
rvc_infer(rvc_index_path, index_rate, vocals_path, output_path, pitch_change, f0_method, cpt, version, net_g, filter_radius, tgt_sr, rms_mix_rate, protect, crepe_hop_length, vc, hubert_model)
File "/content/FIX/src/rvc.py", line 150, in rvc_infer
audio_opt = vc.pipeline(hubert_model, net_g, 0, audio, input_path, times, pitch_change, f0_method, index_path, index_rate, if_f0, filter_radius, tgt_sr, 0, rms_mix_rate, version, protect, crepe_hop_length)
File "/content/FIX/src/vc_infer_pipeline.py", line 571, in pipeline
self.vc(
File "/content/FIX/src/vc_infer_pipeline.py", line 424, in vc
npy = np.sum(big_npy[ix] * np.expand_dims(weight, axis=2), axis=1)
IndexError: index -1 is out of bounds for axis 0 with size 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/content/FIX/src/main.py", line 346, in <module>
cover_path = song_cover_pipeline(args.song_input, rvc_dirname, args.pitch_change, args.keep_files,
File "/content/FIX/src/main.py", line 316, in song_cover_pipeline
raise_exception(str(e), is_webui)
File "/content/FIX/src/main.py", line 85, in raise_exception
raise Exception(error_msg)
Exception: index -1 is out of bounds for axis 0 with size 0
1
forest sluice
anyone know how to fix this ```D:\Program Files\Applio\Applio-3.0.4\env\lib\site...
Ignore the warning, should be fine
silent sun
do i need python to run the application on mac OS?
i try to open the application but won't work unles this is the wrong app
i downloaded the app from the github page
forest sluice
do i need python to run the application on mac OS?
run the startHttps.command (or startHttp.command if it doesn't work)
silent sun
it doen't work both of them
do i need python or somthing?
i can only find window tutorials
no mac
both wont work
rare gobletBOT
Ayo? @silent sun level 2 !!!
forest sluice
Eh, you can't double click the command file?
silent sun
i can double click
tall ocean
I made an RVC model and did the best I could to get a quality result (cleaning up dataset, trying different batch sizes, preventing overfitting w/tensorboard) and it is still a bit disappointing. The dataset that I am using is not long (~9 minutes), is 200 epochs the best that I am probably going to get?
silent sun
it just comes up with 2 command scripts
forest sluice
https://github.com/w-okada/voice-changer/blob/master/README_en.md#1-usage-with-pre-built-binaries
From the guides:
For Mac version, after unzipping the downloaded file, double-click the startHttp.command file corresponding to your VC. If a message indicating that the developer cannot be verified is displayed, please press the control key and click to run it again (or right-click to run it).
proud elbow
I made an RVC model and did the best I could to get a quality result (cleaning u...
go 500 epochs if you use default pretrain
smoky oxide
what feature retrival rate is optimal in order to get the best sounding audio, and which f0 method produces the best results?
I think that will just expand the existing flaws, since it will learn the flaws again
proud elbow
you used the TTS voice as dataset or just for inference?
smoky oxide
My guess is that youre better off uploading the original data set to eleven labs
rare gobletBOT
Ayo? @smoky oxide level 1 !!!
tall ocean
I'll let you know how it goes.
smoky oxide
no defiantly try, I mean, I dont know what will happen lmao
rare gobletBOT
Ayo? @tall ocean level 4 !!!
smoky oxide
lmao, ill hope for you too
random ice
in the ai voice changer how do you actually make it go through the output and not windows default
smoky oxide
when cloning a voice, what feature retrieval rate do you use
random ice
its still coming out of the windows output device if i change it
rare gobletBOT
Ayo? @random ice level 2 !!!
tall ocean
when cloning a voice, what feature retrieval rate do you use
I just leave it at default, you using Mangio?
tall ocean
its still coming out of the windows output device if i change it
did you try checking your actual audio settings?
random ice
did you try checking your actual audio settings?
wdym
tall ocean
smoky oxide
im going to be honest, i have no clue lmao, its pretty janky. it runs through python on my computer, looks like this
random ice
what am i looking for in these settings?
smoky oxide
Ive found the most success though .4 but i haven't tested much
tall ocean
I don't want to make you change what software you are using but I personally use MangioRVC
random ice
what does the mangio one look like
smoky oxide
Honestly i have no attachment to any particular software
im pretty sure im just using a rudimentary interface with access to RVC
tall ocean
random ice
is the mangio fork better than the default one?
tall ocean
is the mangio fork better than the default one?
I don't want to spam the channel with images anymore but here's the page you would be interested in:
random ice
looks pretty much the same to me
tall ocean
im going to be honest, i have no clue lmao, its pretty janky. it runs through py...
as this one?
random ice
as this one?
nah this one
tall ocean
yes but minshin isn't using either of these
random ice
who is minshin
random ice
oh yeah that guy
proud elbow
what's the dataset length? 5-10 mins of natural voice is enough imo
random ice
btw what im trying to say with the ai voice changer the audio is coming out of this one instead of that one
tall ocean
idk restart computer LOL
random ice
ok ill try
proud elbow
btw what im trying to say with the ai voice changer the audio is coming out of t...
use VAC (virtual audio cable) and it doesn't need to reboot, the link is here https://rentry.co/VoiceChangerGuide
tall ocean
use VAC (virtual audio cable) and it doesn't need to reboot, the link is here ht...
yes good idea, immediately though of VAC when I downloaded that voice changer
smoky oxide
between crepe and harvest which f0 method do you find most sucess with
random ice
use VAC (virtual audio cable) and it doesn't need to reboot, the link is here ht...
i downloaded VAC then what i set it as before started working
wtf
old oracle
how many eps should I do for around 45 mins of data?
rare gobletBOT
Ayo? @old oracle level 4 !!!
serene wigeon
is there some place that explains how to use the #🔍│find-models bot? I notice that I search terms with that bot and in #1175430844685484042 I get different results. If I find a model in #1175430844685484042 will that work without needing to find it in #🔍│find-models ?
icy umbra
how do i even use this rvc
rare gobletBOT
Ayo? @icy umbra level 1 !!!
brave garnetBOT
⠀
Google Colabs 
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't inclued 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 💖
molten pecan
is there some place that explains how to use the <#1163592055830880266> bot? I n...
https://aihubdocs.github.io/en/ check the voice models page
There's a tutorial for it + more methods
;3
molten pecan
how do i even use this rvc
Which one exactly?
molten pecan
how many eps should I do for around 45 mins of data?
First, going for 45 mins is prob a lil too much and unnecessary
Second, there isn't a way to know the right epochs amount, as it depends on the length, quality, and size of it
keen peak
i need a 21 savage voice model can anyone help
molten pecan
Third, inputting a semi-arbitrary amount makes you prone to underfitting/overtraining. Will influence the final quality
https://aihubdocs.github.io/en/ use Tensorboard instead. Check the tensorboard guide =3
molten pecan
What dose Ignore outliners in chart scaling do just need a reminder
@fluid matrix
Do you remember?
molten pecan
i need a 21 savage voice model can anyone help
With making an AI cover?
molten pecan
https://aihubdocs.github.io/en/ use Tensorboard instead. Check the tensorboard g...
Check for the tutorial here
azure marshBOT
RVC Guides (How to Make AI Cover)
torn seal
can anybody help me save my model in colab disconect?
i trained it up to 1000 epochs and could not figure out how to save to google drive. like it used to
i dont care about making a cover i was training
long scroll
May I ask how do I use the search voice models feature? I'm trying to find chevruse from genshin
thin stump
May I ask how do I use the search voice models feature? I'm trying to find chevr...
/search model is nice and all but #1175430844685484042 is better since it comes from AI HUB so you can expect quality
dawn nebula
i trained it up to 1000 epochs and could not figure out how to save to google dr...
then u need to train it again
or
do you have sth starts with D_xxxx.pth
and G_xxxx.pth
by any chance
brazen verge
brave garnetBOT
⠀
Download for Nvidia GPUs 
Version 17b cuda
Download for AMD GPUs 
Version 17b directml
Download for Intel GPUs 
Version 17b directml
Download for Mac 
Version 17b Mac
⠀
Made with love by Ilaria 💖
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't inclued 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 💖
long scroll
/search model is nice and all but <#1175430844685484042> is better since it come...
Can you tell me how I can find chevruse (girl from genshin) using the search in #1175430844685484042
rare gobletBOT
Ayo? @long scroll level 2 !!!
thin stump
Can you tell me how I can find chevruse (girl from genshin) using the search in ...
Search for it. I linked it in the #uknown link you saw in my post
long scroll
Search for it. I linked it in the #uknown link you saw in my post
Ohh so that what it was thank you!
brave garnetBOT
Made with love by Ilaria 💖
thin stump
Ohh so that what it was thank you!
Np, I used genshin as the term because sometimes the specific character doesn't show up
long scroll
Np, I used genshin as the term because sometimes the specific character doesn't ...
Boo yah thats what I'm talking about thanks! 10/10
pine ore
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 💖
civic onyx
Ai cover gen is shitting itself
"Traceback (most recent call last):
File "/content/Hina_RVC/src/webui.py", line 8, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'"
@acoustic scarab
nvm it was the token thing that messed it up i think
nope its actually just fuckin broke
its not reading the vocal remover/seperator
hot lagoon
Karaoke vs BVE in uvr?
Wich and why would you use one of these 2
glad zealot
Ai cover gen is shitting itself
did you run all the cell? since ive been using it and that ever never shows up
you can just use this
ngrok is not really needed unless google decides gradio wont work anymore
humble path
-colab
rain pawn
I was learning Rvc with the RVC-beta-v2-0618.7z model in the link above "https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main" and I got an error. It says it can't create the file, but I thought it might be a permissions issue, so I ran it as administrator and still got the error. How can I resolve that error?
rare gobletBOT
Ayo? @rain pawn level 1 !!!
rain pawn
proud elbow
try not to have special characters (non-ASCII) in the path name, also you had the minibatch thing because of 1h+ long dataset, which could significantly reduce the indexing accuracy
rain pawn
The error has been resolved. Thank you for your help.
silent sun
guys i try to open the application on mac os and it won't open how to fix do i need to download somthing?
spiral trench
-download
Does anyone have the link I should be using for a 3070 rtx?
I feel like i downloaded the wrong one
for RVC
tranquil cliff
-rvc
spiral trench
Im looking for the real time voice changer i might have been confused not the one that makes songs.
Would that be rvc2?
rare gobletBOT
Ayo? @spiral trench level 1 !!!
tranquil cliff
ahhh I understand
azure marshBOT
🎙️ Realtime Voice Changer (Local)
tranquil cliff
these are the guides to change voice in real time
azure marshBOT
🎙️ Realtime Voice Changer (Local)
thin stump
I've been seeing this Ov2 (pretrained) in #1175430844685484042. How does this apply to rvc collab and when do you use it?
Nvermind, 6 minutes maximum and it's a setting in the colab
tranquil cliff
I've been seeing this Ov2 (pretrained) in <#1175430844685484042>. How does this ...
what colab do you use?
thin stump
I was planning to use rvc disconnected
tranquil cliff
thin stump
Yup I forgot it was mentioned in the guide here https://aihubdocs.github.io/en/rvc/cloud/training/rvc-disconnected/
brave garnetBOT
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't inclued 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 💖
azure marshBOT
RVC Guides (How to Make AI Cover)
jolly veldt
can i use the web rvc version to talk with the ai voice
and let anyone hear my ai voice?
errant barn
arkadaşlar kendi sesimi nasıl başka sese dönüştürürüm
brittle wing
why i start voice changer and people dont hear my voice, even tho i hear record and it works
errant barn
BRO I'M DONT UNDERSTAND ME
half cove
before asking questions about voice changer please read voice changer guide and if you have any problems after then ask
errant barn
NEREDE BULABİLİRİM ONU
errant barn
THANK YOU BRO
half cove
before asking questions about voice changer please read [voice changer guide](ht...
also for questions about it please use #🔍│help-w-okada
crisp apex
Hello, does anyone have the colab link to train?
half cove
Hello, does anyone have the colab link to train?
-colab
crisp apex
ok
rare gobletBOT
Ayo? @crisp apex level 1 !!!
crisp apex
Ayo? <@726455808933756990> level **1** !!! <:lfg:1159355870119993496>
omaiga
warm beacon
How can i make custom lyrics to be transferred onto original song singing?
proper shale
you need someone to sing those lyrics so you can convert that audio with RVC
warm beacon
could text to speech be used somehow?
celest lantern
pls help me
proper shale
could text to speech be used somehow?
not really, since no tts can sing with a predefined melody
proper shale
pls help me
what's up
fluid matrix
What dose Ignore outliners in chart scaling do just need a reminder
@molten pecan i mean just like it implies, it excludes/ignores really extreme values from the graph to visualize it clearer. Basically smoothing
cosmic tide
This is the rvc that i made?
dreamy lynx
hey i can't seem to downloaded mangio
molten pecan
<@791013455251374151> i mean just like it implies, it excludes/ignores really ex...
I see, ty
dreamy lynx
i click on the link but nothing happens
proper shale
This is the rvc that i made?
yes
dull orchid
Is there a way to fix the problem that i have at the moment, i am using RVC local and when i convert the audio to my model some parts have artifacting or its sounding robotic, mostly on high notes. all other parts sound good
cosmic tide
Now i need to publish it
dull orchid
i did remove a bunch of reverb and echo
proper shale
Is there a way to fix the problem that i have at the moment, i am using RVC loca...
then it's probably bcuz you didnt give sufficient data for those high notes, but ya can transpose the audios u use
aka change pitch
dull orchid
then it's probably bcuz you didnt give sufficient data for those high notes, but...
I think this is 100% correct since i didnt train those high notes, didnt think of that one
and what do you mean with Transpose ?
proper shale
or within RVC itself
RVC itself
dull orchid
Alright
proper shale
and what do you mean with Transpose ?
just another word for changing pitch
proper shale
Also is it better to convert small parts ?
Nah
dull orchid
should i just cut it and do max 30seconds for example ?
Alright
Thanks i will try that right now
hope it will work 😉
preciate you
proper shale
yw :)
dreamy lynx
is mangio the best rvc to use?
azure marshBOT
half cove
original rvc
dreamy lynx
ooo the legendary poopmaster himself
rare gobletBOT
Ayo? @dreamy lynx level 1 !!!
dreamy lynx
yooo
dull orchid
Poopmaster is AICoverGen any good ?
dreamy lynx
wait which one is mainline
dreamy lynx
original rvc
o
half cove
Poopmaster is AICoverGen any good ?
its better to just separate vocal and instrumentals yourself
also its kinda old
dull orchid
Aright will do that 🙂
dreamy lynx
"Original RVC"
yup still thx
dull orchid
By the way will it make a difference if i make a dataset with only singing for a sing model ?
Right now i have a dataset that is only talking and i make it sing trough RVC
I did train that model ages ago, i think i can push way more out of it xD
And how long should my dataset be with singing only ?
10 - 30 mins ? or has that changed over the months
dreamy lynx
original rvc
oh i remember this i tried to use it before but i could only open the real time voice changer couldn't find anything else do you some sort of tutorial on this?
half cove
were you using the amd one?
half cove
they deleted it
dreamy lynx
oh
half cove
i dont even know why
proper shale
By the way will it make a difference if i make a dataset with only singing for a...
will have more range
...depending on the type of singing ofc
talking datasets can sing but have limited range
dreamy lynx
i dont remember if mangio has amd/intel support
i really screwed up by getting a amd
first ai art now ai voice
sand iris
i dont remember if mangio has amd/intel support
never did
half cove
i really screwed up by getting a amd
you can just use colab
dreamy lynx
you can just use colab
oh true i suppose
broken quartz
what is pretrain on RVC V2 disconnected ?
half cove
what is pretrain on RVC V2 disconnected ?
it has original rin_e3 and ov2
half cove
how long should i make the dataset ?
id say like 15-20 min
dull orchid
Alright i will get 15 - 20mins pure only singing
and see if that will change the outputs
compared to a dataset with only talking
broken quartz
it has original rin_e3 and ov2
I saw, what is it for?
cosmic tide
After i download that, what i need to do?
proper shale
how long should i make the dataset ?
try going for 10+ mins
half cove
I saw, what is it for?
its just the pretrains it uses while training id recommend ov2 for datasets under 5min and original for anything longer
proper shale
After i download that, what i need to do?
u need to zip the .pth with ur model's name (G/D are training files, not needed) and the index together
broken quartz
its just the pretrains it uses while training id recommend ov2 for datasets unde...
ok thanks
cosmic tide
u need to zip the .pth with ur model's name (G/D are training files, not needed)...
So just zip it?
Like that?
cosmic tide
And what is that?
proper shale
And what is that?
that's for training, just the preprocessed files and logs
dull orchid
Should i add singing with a bit of reverb into the dataset or keep it clean and raw vocals only ?
rare gobletBOT
Ayo? @dull orchid level 3 !!!
dull orchid
not sure if that would be better or not
proper shale
I can publish it?
yeah, if you've done what I've said u should be good
proper shale
Should i add singing with a bit of reverb into the dataset or keep it clean and ...
u can de reverb vocals yk
dull orchid
u can de reverb vocals yk
Yeah ? alright i add vocals with reverb as well 😄
proper shale
It's different from that
npy file isnt needed
cosmic tide
No
rare gobletBOT
Ayo? @cosmic tide level 5 !!!
proper shale
This is the folder
then u screwed up on training the model
proper shale
...yeah
cosmic tide
Ok
sour dune
Can anyone help me? whenever i run the start bat file a error pops out it's saying "Error loading python dll"
oak barn
I'm having trouble making the cover? Can someone dm me?
dull orchid
Got 21 minutes of singing for my dataset 😄
with clean raw rocals some vocals with reverb
lets see how the outcome will be
proud elbow
Can my pc train a model ?
of course, can train with batch size up to 6-8
dull orchid
Alright first time training a model again since a long time lol
is there a guide ?
Target sample rate: 40K or 48K ?
Version V1 or V2 ?
proud elbow
with clean raw rocals some vocals with reverb
reverb should never be your friend in datasetting
dull orchid
Awhhh i asked that and someone told me that would be fine
now i have to clean up 21mins of audio xD
dayumm
i knew it had to be clean vocals
proper shale
Target sample rate: 40K or 48K ?
32k
proper shale
Version V1 or V2 ?
v2
proper shale
is there a guide ?
proper shale
no option for 32k
set version to 2
proud elbow
Target sample rate: 40K or 48K ?
depend on frequency cutout in spectrogram
proper shale
and it should appear ;)
proper shale
depend on frequency cutout in spectrogram
nuh uh
never use anything higher than 32k 
dull orchid
is 32k best yeah ?
proper shale
is 32k best yeah ?
its like... ur just adding more noise with 40-48k
dull orchid
alright and you told me i can use a dataset with a bit of reverb vocals right 😮 ?
proper shale
most mics have a cut off at 15khz, which is 30k
proper shale
alright and you told me i can use a dataset with a bit of reverb vocals right 😮...
nope
you can dereverb them tho
dull orchid
yeah i will fix that soon
dull orchid
you can dereverb them tho
what de reverb you are using ?
proud elbow
nuh uh
never use anything higher than 32k <:trolley:1159468147133395025>
most dataset I had are in 16k cutout, so I always went 32k
proper shale
what de reverb you are using ?
foxjoy's dereverb model, but u can do UVR's
proud elbow
foxjoy dereverb and then uvr deecho
dull orchid
Alright thanks both
should i train a feature index first or at what point you do that again
odd obsidian
hii, how to search models? it says did not respond
dawn nebula
hii, how to search models? it says did not respond
which bot
proper shale
foxjoy dereverb and then uvr deecho
ehhhh de echo is not needed 90% of the time
proper shale
should i train a feature index first or at what point you do that again
after preprocessing yh
and feature extraction
but before training
dull orchid
and feature extraction
alright so you do these 2 steps first and than you start training ?
dawn nebula
show me a ss of bot broke
dull orchid
so i understand it correct
proud elbow
ehhhh de echo is not needed 90% of the time
depend on what kind of reverb, in my case it's with literally echo
proper shale
alright so you do these 2 steps first and than you start training ?
yeah, do preprocessing, feature extraction, train index and then training
proper shale
depend on what kind of reverb, in my case it's with literally echo
oh that's fair then
odd obsidian
oh its modelAI
rare gobletBOT
Ayo? @odd obsidian level 1 !!!
dull orchid
another question lol, how many epochs should i do with about 18 - 21min dataset
proper shale
another question lol, how many epochs should i do with about 18 - 21min dataset
go for 500, every model
dull orchid
decent to good quality singing vocals
proper shale
go for 500, every model
if not overtrained ---> go for 750, 1000, etc
if it overtrains before, just stop training ;)
proud elbow
another question lol, how many epochs should i do with about 18 - 21min dataset
500 for default pretrain, or 300 for ov2super, for any dataset length personally...
dull orchid
if it overtrains before, just stop training ;)
Alright and you can see that in the Tensorboard right ?
dull orchid
500 for default pretrain, or 300 for ov2super, for any dataset length personally...
Can you explain what ov2super is ?
im a little rusty when it comes to training
been a while :X
had all my models i needed
proper shale
Alright and you can see that in the Tensorboard right ?
yup
proper shale
Can you explain what ov2super is ?
new pretrains
tbh i would not really recommend using them rn, wait until ya get used to training and then maybe try that
it is better for small datasets (0-10 mins) overall
proud elbow
Can you explain what ov2super is ?
a new pretrain, check it out for 32k https://huggingface.co/poiqazwsx/Ov2Super32kfix/tree/main
dull orchid
Alright got it, going to eat dinner and start training, not sure how long a model will even take to fininsh on this pc 😮
proud elbow
rvc disconnected also has ov2 option tho
dull orchid
when i trained models that was still on Colab back when it used to work xD
dull orchid
a new pretrain, check it out for 32k https://huggingface.co/poiqazwsx/Ov2Super32...
Thanks 😄
half cove
a new pretrain, check it out for 32k https://huggingface.co/poiqazwsx/Ov2Super32...
huggingface is goofy so i cant see how many people have downloaded this 
it always shows 0
smh
odd obsidian
show me a ss of bot broke
proud elbow
huggingface is goofy so i cant see how many people have downloaded this <:boohoo...
I used it for my recent models tho (ofc the fixed one)
half cove
I used it for my recent models tho (ofc the fixed one)
yea i know people have downloaded it
its literally downloaded anytime anyone uses colab but huggingface is poopy
glad zealot
yea i know people have downloaded it
its literally downloaded anytime anyone use...
i probaly already downlaoded that with kaggle around 10x+
dawn nebula
use applio bot
glad zealot
then another 10+ on colab
half cove
i probaly already downlaoded that with kaggle around 10x+
its prob been used thousands of times
odd obsidian
use applio bot
what to write in the technology?
molten pecan
what to write in the technology?
@brittle wing the username
"Forsaken me"
A lil familiar, ain't it?
brittle wing
Yes, Jesus shouted it
odd obsidian
<@456226577798135808> the username
hmm thats what im using in othr games
molten pecan
hmm thats what im using in othr games
I like it
molten pecan
what to write in the technology?
If you are not going to use kits.ai, write RVC
odd obsidian
If you are not going to use kits.ai, write RVC
im looking for gura model
dawn nebula
what to write in the technology?
rvc
brittle wing
How many minuites of datasets for a human like model???
mortal delta
So i need help to find a software that can make AI Covers whitout paywall, and get AI models for Lucas Lee and Gideon Graves.
rare gobletBOT
Ayo? @mortal delta level 1 !!!
cosmic tide
This is the rvc?
lusty willow
Hey there everybody! Currently trying to get to grips with the new easyGUI colab. I appear to be having issues with downloading and using voice models from huggingface? Either the model doesn't download or I get greeted with an error when I attempt to use it.
half cove
**How many minuites of datasets for a human like model???**
id use around 15-20 min
dataset quality matters more
half cove
This is the rvc?
you seem to have downloaded a model and venv bat but your not really telling me what your trying to do
half cove
are you trying to use fcpe on mainline realtime?
dull orchid
Is it still needed to cut the audio of the dataset in 5 - 10 seconds ?
cosmic tide
you seem to have downloaded a model and venv bat but your not really telling me ...
Im trying to make voice model
cobalt saddle
are you trying to use fcpe on mainline realtime?
I just wanted to make ai cover i have model done already
half cove
Im trying to make voice model
im guessing your doing it locally what fork are you using
half cove
I just wanted to make ai cover i have model done already
do go-web.bat
not go-realtime.bat
cosmic tide
im guessing your doing it locally what fork are you using
Colab
half cove
Colab
create your dataset then make your model in colab disconnected https://aihubdocs.github.io/en/rvc/cloud/training/rvc-disconnected/
azure marshBOT
RVC Guides (How to Make AI Cover)
cosmic tide
create your dataset then make your model in colab disconnected <https://aihubdoc...
What is dataset? The audio?
half cove
So i need help to find a software that can make AI Covers whitout paywall, and g...
use rvc you can do it locally (on your pc) or on the cloud (using google colab)
half cove
What is dataset? The audio?
yep
dull orchid
poopmaster :X
cosmic tide
I just need to make with the audio the voice model
half cove
do you need the dataset to be in 5 - 10 seconds parts ?
it does not need to be separated
half cove
or it can be 5 minutes long
yes
half cove
I just need to make with the audio the voice model
then read the guide i sent prior
dull orchid
was looking at a tutorial from 8 months or so ago xD
cosmic tide
then read the guide i sent prior
Ok
cosmic tide
then read the guide i sent prior
I followed it
half cove
elaborate
cosmic tide
# RVC Guides (How to Make AI Cover)
Are you talking about this thing?
half cove
no
dull orchid
For feature extraction do i pick rmpve or rmvpe_gpu ?
rare gobletBOT
Ayo? @dull orchid level 4 !!!
dull orchid
wich one has the best result
half cove
For feature extraction do i pick rmpve or rmvpe_gpu ?
rmvpe_gpu
frozen osprey
hi, if the file is rar and I covert it to zip will simply work like that on google colab?
like the voice model
proper shale
hi, if the file is rar and I covert it to zip will simply work like that on goog...
rar can break stuff, so no
frozen osprey
oh ok, thx
dull orchid
Batch size per GPU: 4 or 6 with a RTX 3060 TI 8 GB ?
what is a good number to start with
i have gone with 5 just to start and see how it goes
proper shale
Batch size per GPU: 4 or 6 with a RTX 3060 TI 8 GB ?
6 is prolly fine
proper shale
i have gone with 5 just to start and see how it goes
eh ok
dull orchid
Yeah should i go with 6 next time ?
proper shale
if it aint broke... dont fix it
proper shale
Yeah should i go with 6 next time ?
u can test
dull orchid
wait batch size per GPU means higher is faster done training ?
dull orchid
if it aint broke... dont fix it
First time training local :X
proper shale
wait batch size per GPU means higher is faster done training ?
yes, but it can affect quality if too high
proper shale
so like, stick with 4-8
dull orchid
Will do
proper shale
I'd say 6/8 are perfect for 90% of models ;)
proper shale
Sounds good, isnt it a problem that i did 5 than xD ?
Eh... it's not that bad lol
Actually that's a good finetune number, when you find the right batch number for your model 😭
proper shale
Actually that's a good finetune number, when you find the right batch number for...
this is done with a lot of testing tho
dull orchid
Yeah i will have to try things out again, Epoch 34 at the moment, so i will watch a movie and come back to my pc
going for 500 epochs like you both said 😄
one more thing, where can i see the Tensorboard
half cove
one more thing, where can i see the Tensorboard
put this in your rvc folder and run it
dawn nebula
lol tensorvenv
proper shale
put this in your rvc folder and run it
you gotta run this every time you wanna open TensorBoard btw
brittle wing
can someone train a voice model for me
dawn nebula
u pay people then
brittle wing
oh ok
dawn nebula
or if someones kind enough they gonna train
dawn nebula
oh ok
do a paid request or #1191429836321849435
torn seal
so with the rvc disconect. https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS i saved the model and everything but how do i save the weights folder?
gloomy quartz
I can’t send pics here but whenever I open rvc I don’t see any of the voices things like it’s hard to explain cuz u don’t have photo perms but all it says is server control and more stuff
torn seal
im trying to save the weights folder to my google drive. can anybody help me with this?
tall ocean
drag and drop right?
torn seal
can you drag and drop from collab?
tall ocean
I had a question too but now don't want to put a huge picture that essentially blocks out your question... lol
torn seal
i found a prompt to using the collab ai but step 2 dont upload the ziped weights folder to my google drive.
whats you question
tall ocean
I trained for 1000 epochs. But it isn't actually worth using the 1000 epochs right? I should use whatever is at 60k.
torn seal
so i heard thats what u are supossed to do
tall ocean
Alright
torn seal
yea i train to 1000
sometimes i go for the 500 or 600 epoch
just depends i think, thats why i wanted the weights folder and listen to each one and pick the best one
tall ocean
when people post their RVC models at 1000 epochs, I wonder if they are stopping it when the sample rate is optimal or 1000 epochs no matter what
The best I got was 390
torn seal
i dont think it matters
390 is just the number of times it has been trained
you can always resume it
rare gobletBOT
Ayo? @tall ocean level 5 !!!
torn seal
kinda sucks i can import the weights folder to my google drive
torn seal
they removed auto save to google drive
so you have to do everything mannually or you will loose the files
tall ocean
want me to train something for you?
torn seal
no i have the model already
rare gobletBOT
Ayo? @torn seal level 7 !!!
torn seal
its just the final model tho at 500 epoch
tall ocean
ah got it
torn seal
i was hoping to get the rest of them
tall ocean
yoo I HAVE A LOT OF DELAY is there any way to fix?=
Realtime voice changer client?
These are the settings that I use
torn seal
but if you have any advise on cleaning up vocals from say a podcast or recording live from youtube be nice. sometimes there is echo in the recordings from his iphone, and there is music behind his podcast on youtube wich has the best audio qaulity of voice but the damn music behind it.
thats cool d1
tall ocean
but if you have any advise on cleaning up vocals from say a podcast or recording...
Hmm maybe we can try using RX10 on it?
tranquil cliff
Hi
tall ocean
if you send me the problematic audio I'll see if I can fix it for you
torn seal
alright add me?
half cove
Hmm maybe we can try using RX10 on it?
if its echo id use uvr
torn seal
hi poop
half cove
spectral denoise is the goat rhe
tall ocean
if its echo id use uvr
this guy knows what he's talking about
torn seal
so do you know how to get the weights folder using https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS
tall ocean
I have UVR as well, wanna send me the audio?
half cove
so do you know how to get the weights folder using https://colab.research.google...
with that colab you can only download one checkpoint at a time
you could
try
torn seal
how should i send it?'
jolly bone
Realtime voice changer client?
yes
rare gobletBOT
Ayo? @jolly bone level 1 !!!
half cove
downloading it from the left
jolly bone
I sent you some settings
i'll try that cuz i want to be a girl in game but im a man
half cove
because i just wana be able to pick and choose get it over with
just download a few to test
to make sure you havent overtrained
torn seal
fuck thats so gay
so there is a way to download the weights folder tho using the file explorer?
half cove
with colab u cant download folders but u can download everything inside
torn seal
ok how can i do that?
half cove
ok how can i do that?
just on colab go on the left file manager then mangio then weights open folder and download everything inside
takes awhile tho
jolly bone
I sent you some settings
it still have delay
torn seal
ok thanks for the info poop, so if i right click there is no option to download. is there a botton i press instead?
gloomy quartz
I have bad delay and it dosent really pick up my mic well it just cuts out a lot and dosent pick up my word even if I yell can anyone send me something
half cove
ok thanks for the info poop, so if i right click there is no option to download....
3 dots then download but u can just use this
torn seal
so im testing my model. is it possible to just through in the entire acapella to convert my model to, its the whole 3 min song. or do i have to cut it up ?
dawn nebula
what
patent trellis
half cove
dawn nebula
good
dawn nebula
good
torn seal
im thinking the pitch needs one down tho, wish i can show u guys
if you guys are familur with nopoleon from the outlawz sang with 2pac, turned muslim, he dont rap no more.
i havnt tryed training it with rin_e3 yet, i chose the ovr or what ever its called. my datasert is not 10 min but like 5 min, any adivese on wich pretrained i should use?
dawn nebula
i havnt tryed training it with rin_e3 yet, i chose the ovr or what ever its call...
ov2
torn seal
yea thats the one ov2
brittle wing
dawn nebula
你好像没装python
brittle wing
有阿
torn seal
i heard one of those pretrained models are better for shorter datasets, does anybody know
?
dawn nebula
i heard one of those pretrained models are better for shorter datasets, does any...
yeah thats ov2
dawn nebula
有阿
https://www.python.org/downloads/release/python-390/ 再装一遍，记得把path选项勾上
torn seal
for real like whats the shortest delik?
brittle wing
https://www.python.org/downloads/release/python-390/ 再装一遍，记得把path选项勾上
那我舊版本要刪掉嗎
rare gobletBOT
Ayo? @brittle wing level 2 !!!
torn seal
im almost tempted to go 1000 epochs i only went to 500 with 5 min dataset, high qaulity. or would that be over kill?
molten pecan
me and char revamped the troubleshoot command
@fluid matrix reminder for you too
brittle wing
這樣應該是成功了吧
brittle wing
還是不行
dawn nebula
https://tenor.com/view/sure-john-cena-are-you-sure-gif-5783187
wait
brittle wing
是還要再修改路徑嗎
brittle wing
我英文有點問題 也看不太懂跟我有相關疑問的人是如何解決的
sand iris
@dawn nebula sorry for irrupting this way, but when using a fat zip of the Original RVC, you don't need to install python
dawn nebula
<@690852174267416646> sorry for irrupting this way, but when using a fat zip of ...
yeah mb
sand iris
It comes provided with everything else
dawn nebula
how come this person cannot use the thing
sand iris
I think this happens when extracting with anything but 7zip
brittle wing
<@690852174267416646> sorry for irrupting this way, but when using a fat zip of ...
ok
sand iris
WinRAR & the windows default thing have been buggy before
dawn nebula
@brittle wing 你用什么解压的
sand iris
7zip should work
brittle wing
<@456226577798135808> 你用什么解压的
就文件管理員
brittle wing
WinRAR & the windows default thing have been buggy before
so i should use which tools
gloomy quartz
Why is no one helping me 😭😭
rare gobletBOT
Ayo? @gloomy quartz level 1 !!!
sand iris
brittle wing
好的 謝謝 我看那個外國老哥 是說不用安裝python嗎
dawn nebula
如果点右边的下载就会装很多别的东西
dawn nebula
delik do you guys have manderin keyboards
yeah im chinese lol
torn seal
like how do you guys type like that?
dawn nebula
type english letters then
torn seal
whats your keyboard look like
torn seal
krazy
brittle wing
rare gobletBOT
Ayo? @brittle wing level 3 !!!
brittle wing
這樣可以了嗎
torn seal
so each each box a word or letter?
brittle wing
謝哥 成功了
torn seal
so how do you translate english keyboard to a character?
dawn nebula
没事，you should thank @sand iris lol
dawn nebula
so how do you translate english keyboard to a character?
pinyin
brittle wing
https://www.7-zip.org/download.html
thanks for your help
torn seal
pinyin ? turns into a character
brittle wing
I succeeded
torn seal
oh i get it, so you type what your saying in the alphabet and it converts it
dawn nebula
yeah
torn seal
thats cool. im chicano, in america
dawn nebula
you dont need to type full ones, it can auto detect
torn seal
i had a homie that was from tiawan
his name was rick or at least thats what he says his parents named him, from some cartoon that is popular there,
he was an odd one, i loved picking on him. haha
brittle wing
我該如何選擇
half cove
fyi if you click v1 then v2 then 32k will magically appear
torn seal
are you trying to teach that fella how to train?
half cove
fyi if you click v1 then v2 then 32k will magically appear
@dawn nebula translate this
i dont wanna get it wrong
dawn nebula
<@690852174267416646> translate this
done
half cove
i was heating up my breakfast
dawn nebula
ilaria rvc on colab
dawn nebula
直接提取特征就行
proper shale
dawn nebula
bro u faster gg
rare gobletBOT
Ayo? @brittle wing level 1 !!!
dawn nebula
https://aihubdocs.github.io/en
there is a section for ilaria rvc
brittle wing
你的顯卡是哪一張
cobalt saddle
do go-web.bat
this is what i have
half cove
this is what i have
is this amd/intel?
cobalt saddle
is this amd/intel?
half cove
what
cobalt saddle
<@396054102938877952> u training or making covers?
i make cover right now
half cove
i make cover right now
then use that
for weights.gg models you need to reupload them as they need auth
for weights.gg models to use on ila rvc you need to reupload it
you can also
upload it to google drive
make it public
get the share link
and do that
cobalt saddle
<https://aihubdocs.github.io/en/essentials/how-to-upload-models-to-hugging-face/...
should i just put my model link from huggingface?
half cove
should i just put my model link from huggingface?
on the download voices tab
at the top
cobalt saddle
oh alr
half cove
<@936673139419664414> what does training voice do
training voice is to make your own voice model
cobalt saddle
if i have existing model how do i make it more realistic tho
brave summit
how to place datasets in huggin face?
half cove
how to place datasets in huggin face?
you dont put datasets in huggingface
brave summit
you dont put datasets in huggingface <:nails:1159569314848972891>
i mean that
brave summit
i guess
half cove
i think there should be a folder called audios or dataset
applio updated so you should use that
or just mainline
dull orchid
My training is done but i dont see the model ? 😮
half cove
My training is done but i dont see the model ? 😮
where are you checking
half cove
check weights
dull orchid
check weights
Yeah indeed there it is 😄
you can change the index name right >?
its very long now
thin stump
bro copyright
sand iris
bro copyright
Good call lol
craggy brook
What!, is there copyright in this song?
spare shale
**What**!, is there copyright in this song?
Send only vocals
dawn nebula
**What**!, is there copyright in this song?
never share instrumentals
craggy brook
this
snow veldt
Ignore the warning, should be fine
still the best answer ive got, really the only one that even begins to answer it head on really..
what do you know about it, DOES it actually slow people down?
anyone else here running a 4gb card, trying to get a gague on how fast an epoch should train
ripe void
is it a difficult task to get RVC up and running?
austere cairn
So many different RVC versions, are they all pretty much the same? Looking to run on a Linux WSL install, not even planning to train anything, just download a model and mod audio
snow veldt
So many different RVC versions, are they all pretty much the same? Looking to ru...
Replay if you just want to mess with audio and models
on a clean gui
short pecan
hey rq how does one re dub audio with a voice model for Mangio RVC? never done that yet
Been only using it for the voice changer so far but I wanna know how to do audio files too
tired torrent
help
austere cairn
That'll let me download a model from here and change my pre-recorded audio?
snow veldt
That'll let me download a model from here and change my pre-recorded audio?
https://www.tryreplay.io/
yeah bud, its a nice program, can do batches and all sorts, all but train really
tired torrent
help
austere cairn
Cool, thank you!
snow veldt
np!
sand iris
still the best answer ive got, really the only one that even begins to answer it...
It's a feature at this point lol
pretty much always been a thing and never fixed
So not sure anyone would have an answer
sand iris
anyone else here running a 4gb card, trying to get a gague on how fast an epoch ...
it won't fit within 4GB, and unless crash or fallback to system RAM which is tremendously slower
Might want to use colab for training instead
snow veldt
It's a feature at this point lol
hahah, yeah thats come to be my understanding of it xD
snow veldt
So not sure anyone would have an answer
yeah they don't seem to over on AI Hisp
snow veldt
Might want to use colab for training instead
not an option currently, dont have the net for it
it works, but slow like you said, 20 mins an E
for a 16min data set
sand iris
it works, but slow like you said, 20 mins an E
I don't have any proper numbers to give
But yeah you'll be in that kind of range
snow veldt
so that's expected, i can deal with that, its an old rig i only expect to get so much out of
just wanted to know i hadnt done something to make it that.. extra slow when i saw most people are doing an ep in like 30 seconds haha
i think what i should stick to is just doing up the datasets for anyone interested in training some voices we really should have by now, like Scott Weiland
sand iris
You'd have to upload them, if network is an issue
lethal robin
I am here lol
lethal robin
I want to download songs on a mp3 file to extract the music and volcals
snow veldt
You'd have to upload them, if network is an issue <:nails:1159569314848972891>
yeah, one dataset beats a dataset and a D and G file, an index and anything else tho
lethal robin
oh cover lol
half cove
oh cover lol
so are you trying to separate instalments and vocals?
snow veldt
I want to download songs on a mp3 file to extract the music and volcals
replay can do that for you if you're not worried about Quality. straight into the program, and it will split and rejoin for you once its altered the vocals
rare gobletBOT
Ayo? @snow veldt level 5 !!!
lethal robin
Well like i don't even have the songs or music in the first place to separate them in the programs
snow veldt
well with Replay, you can just provide youtube links if you like, ive done a few songs and it does the job, it might not be perfect in regards to bitrate, but it should sound good enough most of the time, if the idea is to just belt out a bunch of stuff or just have some fun playing around, its the way id go
brittle wing
It says "ModuleNotFoundError: No module named 'pydantic' when trying to open the start_http. bat
pls help 😭
😭
i have Nvidia Geforce GTX and AMD processor, idk if thats important. i downloaded python, and i think i installed that 'pydantic' thing but idk e.e. it still wont open
lethal robin
okay cool thanks man i appreciate that and all of your help
half cove
Well like i don't even have the songs or music in the first place to separate t...
you have to get audio to convert first
snow veldt
torrents are not just for pirates..
snow veldt
hundreds of thousands of legit songs and files are shared that way, BradSucks shares all their stuff that way
half cove
server will get in trouble for stuff like that
half cove
hundreds of thousands of legit songs and files are shared that way, BradSucks sh...
you think discord cares about that tho?
snow veldt
that's crazy, Discord have really gone the route of.. lets not even put words on it lol, its just soft
half cove
they nuked our asses last time
snow veldt
if youtube didn't exist, id probably get why discord give a shit. sucks you guys cant just have a nice thing with running a server freely
half cove
so im just not taking any chances
snow veldt
yeah i get hat
that&
omg lol
does anyone actually keep stuff underground tho, databases of converted stuff, vocal ai covers, any of that, cuz its becoming a joke not being able to talk about it or share the stuff anywhere but youtube, i want in on the convos lol
half cove
on youtube aswell they are cracking down on ai lmfao
brittle wing
It says "ModuleNotFoundError: No module named 'pydantic' when trying to open the...
DAREKA TASUKETEEEEEE
rare gobletBOT
Ayo? @brittle wing level 1 !!!
snow veldt
so like 3% of the problem on there? what about the album uploads ive been listening to for 19 years, taking away from the artists
sand iris
I doubt someone listening to a cover would listen to that instead of the original thing tbh
tacit grail
How do you refine an ai cover? You know where it sounds like its mumbling/robotic or something despite using a ton of data of a character from the entire show?
sand iris
Like most of the time I'm seeing covers it's "aha lmao", but I don't necessarily care about the original media
snow veldt
I doubt someone listening to a cover would listen to that *instead of* the origi...
no im talking legit album uploads from some dude, youtube any band, you can probably find stuff of theirs on youtube from over 15 years ago, they never solve a single issue on that platform, unless it makes them big money
sand iris
¯_(ツ)_/¯
snow veldt
mad jump in karaoke's now too, yeah i don't think you're wrong, sounds like youtube to be that tunnelvisioned, thinking ai covers are their problem lol
brittle wing
Hello, does anyone know who the most brainy users are regarding RVC? I need help opening the program for the first time :T
lethal robin
Crazy lol but it was never my intention to use copyrighted material for my own financial gain rather for my friends and my own amusement lol
ripe void
ok so
i have the RVC webui open
an index file and a pth file
how do actually use those files with the RVC webui
keen peak
it says i dont have permition to send messages in the voice models channel
rare gobletBOT
Ayo? @keen peak level 1 !!!
half cove
then it will appear on the dropdowns
ripe void
then it will appear on the dropdowns
epic
keen peak
<https://aihubdocs.github.io/en/essentials/how-to-get-model-maker-role-in-ai-hub...
do i have to make my own?
half cove
do i have to make my own?
to get model maker ofc
keen peak
to get model maker ofc
im trying to get ones that are already made
half cove
im trying to get ones that are already made
then why do you need to send messages in voice models?
cobalt saddle
any tip to make my model not emotionless
keen peak
then why do you need to send messages in voice models?
wait which channel do you use for getting premade models?
half cove
wait which channel do you use for getting premade models?
just download the model
you can see the models just not send messages
half cove
any tip to make my model not emotionless
use audio with more emotion
keen peak
just download the model
how do i search for them tho
half cove
how do i search for them tho
fervent jetty
how do i make an ai voice model
rare gobletBOT
Ayo? @fervent jetty level 1 !!!
glossy tartan
how many epochs do u guys recommend? i usually just set it to 1000 but idk if thats wasting time
austere cairn
So downloaded Replay and downloaded some models from this hub, but it's not sounding so great. What settings have largest impact on results? Stem and F0 method?
young halo
Yo, what is ov2 super?
half cove
Yo, what is ov2 super?
new pretrain
half cove
how many epochs do u guys recommend? i usually just set it to 1000 but idk if th...
i test to do 500 and then more if it hasnt overtrained
half cove
So downloaded Replay and downloaded some models from this hub, but it's not soun...
use rmvpe and stem on replay i dont really know
young halo
how many epochs do u guys recommend? i usually just set it to 1000 but idk if th...
There's not an exactly recommend amount, 1000 is okay, but make sure to use tensorboard to check if it's overtraining or no
glossy tartan
There's not an exactly recommend amount, 1000 is okay, but make sure to use tens...
what is tensorboard?😅
young halo
There's not an exactly recommend amount, 1000 is okay, but make sure to use tens...
I usually stop the training at around 300 or 400 Epochs, but It depends if they're before or past the overtraining point
dawn nebula
what is tensorboard?😅
a place to check how good the model is quantitatively
young halo
what is tensorboard?😅
Do you use RVC Disconnected?
austere cairn
use rmvpe and stem on replay i dont really know
Is a different application preferable to Replay for getting better quality results?
rare gobletBOT
Ayo? @austere cairn level 1 !!!
glossy tartan
Do you use RVC Disconnected?
to make the models i use rvc webui
half cove
Is a different application preferable to Replay for getting better quality resul...
you could just use mainline
-links
azure marshBOT
half cove
original rvc
young halo
to make the models i use rvc webui
I recommend using RVC Disconnected, since it can avoid having interruptions when training a model
half cove
I recommend using RVC Disconnected, since it can avoid having interruptions when...
if you can do it locally id recommend using that instead
austere cairn
you could just use mainline
"Original RVC, by the RVC-Project team GitHub" would this be better?
young halo
I recommend using RVC Disconnected, since it can avoid having interruptions when...
However tensorboard is an orange tab that usually appears when you start training a model, it shows some graphics that help you monitoring your trainings
