try to launch replay with powershell $env:CUDA_LAUNCH_BLOCKING="1"
#✨│ai-help
boreal creek
gpt 5.5 says :If they want the fastest practical fixes, I’d say:
Reduce batch size to 1.
Lower resolution/model size.
Restart PC.
Update NVIDIA Studio or Game Ready driver.
Reinstall Replay/PyTorch environment if it still fails.
Try WSL2/Linux if Windows keeps producing CUDA launch failures.
it could be issues with lack of vram or maybe you are using too much audio?
simple ore
your theory is weird
the goal of the pretrain training is to make a model capable of predicting how a spectrogram would look like for a given speaker, phoneme, pitch
properly trained pretrain should be able to infer (phonemes + pitch) from original audio + speaker vector into the original audio close enough
and it should be able to make a good guess of how a different unseen audio would sound if it was said by a known speaker
when you finetune a pretrain using a new speaker it realigns into predicting how any input would sound with the new speaker's voice
chatgpt's take
generally what happens is that the model overfits and loses ability to predict anything but the content of the finetuning dataset
gray dagger
moving replay to another drive seemed to do the trick
uneven tendon
program works fine with default voices, whenever I upload a custom voice, and swap to it they dont work then the client freezes, if anyone knows this issue? no error in cmd, just ends with...
[Voice Changer] Loading index...
Try loading... model_dir\5\added_IVF256_Flat_nprobe_1_CyreneAidenDawnHSR_v2.index
Version:
MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.18a.zip
viral mason
program works fine with default voices, whenever I upload a custom voice, and sw...
what you're using is extremely old
default voices only come with very old version of the voice changer
fringe heron
generally what happens is that the model overfits and loses ability to predict a...
For me no matter the steps it always gives the "overfitting" results
Predicting b to b right but not a to b reliably
low shard
program works fine with default voices, whenever I upload a custom voice, and sw...
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
viral mason
What voice model are you using, what gpu do you have? (Nvidia or AMD) and what are you wanting to do with each
small sorrel
do u have to upload your model to hugging face
low shard
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
Do you really need to use docker?
hallow thistle
do u have to upload your model to hugging face
Sure.
placid cradle
weights Can't you use it?
light ermine
chat i need a repo id which supports text generation
where do i find it
im completely new to ai stuff btw
low shard
weights Can't you use it?
weights is dead
cosmic ravine
what voice changer do i use
hallow thistle
what voice changer do i use
What is your PC GPU? And what do you use the voice changer for?
cosmic ravine
What is your PC GPU? And what do you use the voice changer for?
its rtx 4050 6gb i wanna make my mic sound bad and crispy
hallow thistle
its rtx 4050 6gb i wanna make my mic sound bad and crispy
For simple microphone audio effect, you could use a simple voice mod program. However, Vonovox and Tg Develop's W-Okada are AI voice changers, have FX options, but Vonovox gives better base audio quality.
cosmic ravine
how to get vonovox
hallow thistle
Download Vonovox. https://huggingface.co/dr87/vonovox/resolve/main/Vonovox_beta_17_11.zip
stark ether
Hola, como hacer un audio con la ia de bad Bunny 2022 UVST
cosmic ravine
Download Vonovox. https://huggingface.co/dr87/vonovox/resolve/main/Vonovox_beta_...
thanks
hallow thistle
Hola, como hacer un audio con la ia de bad Bunny 2022 UVST
What is this Bad Bunny about?
hallow thistle
This is the guide about Vonovox. https://docs.aihub.gg/realtime-voice-changer/local/vonovox/
hallow thistle
Use WinRAR or 7-Zip to extract the zip content. Inside Vonovox folder, double click on start.bat to launch the program.
cosmic ravine
Use WinRAR or 7-Zip to extract the zip content. Inside Vonovox folder, double cl...
what about file explorer
hallow thistle
cosmic ravine
i will be back when it done extract
devout dome
hey is there a way to setup playback on vonovox?
devout dome
what do i do\
when it was my first time opening it took forever but after like 1 12min vid it started working
low shard
its rtx 4050 6gb i wanna make my mic sound bad and crispy
are you looking to make it bad for a simple effect or make like voices more realistic?
low shard
when it was my first time opening it took forever but after like 1 12min vid it ...
youtube videos are outdated asf for RVC related stuff, don't suggest them, what did you use?
devout dome
no no i used a random vid to pass the time while it opened
the tutorals i used were from him its working fine now i just dunno how to setup playback for the vc
like rn im using the 17_11 vonovox beta
low shard
the tutorals i used were from him its working fine now i just dunno how to setup...
please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
devout dome
i meant "here" but it was one Your_Local_Worm sent on the discord.
Nvidia 5070
Win 11
Roleplay/jokin with friends
sorry* im still a lil tired so i keep messing up my words
north drift
Guys, can you help me pls?
[Voice Changer] warming up... generating sola buffer.
got this thing and nothing coming after
low shard
Guys, can you help me pls?
[Voice Changer] warming up... generating sola buffer....
This is a General AI Discord Server, please elaborate:
- your pc gpu
- your pc os
- what are you trying to do: TTS, AI Covers, E Girl Trolling / Catfishing or Roleplay
- the tutorial link used
viral mason
This is much better looking, good job
low shard
This is much better looking, good job
thx, I merged all the help commands into this, we had the forgotten howtoask and an help template lol
viral mason
Lol
pliant stream
That does look good
low shard
That does look good
thx, btw I was a mod in weights too and don't remember you there (I was more active in AI HUB so maybe it's that), but also thanks for the recent help
I think that Weights role might be deleted one day 😭
viral mason
Maybe could keep it as nostalgia
low shard
Maybe could keep it as nostalgia
just checked and it's just a cosmetic role, so there isn't the risk that one day one of their account gets hacked and that it gets abused lol
pliant stream
I think that Weights role might be deleted one day 😭
My role is the greatest thing ever lmao
I honestly dont remember who gave it to me but somebody did
viral mason
just checked and it's just a cosmetic role, so there isn't the risk that one day...
Alr lol, but any links referring to the site should be removed besides a download to replay at least
low shard
My role is the greatest thing ever lmao
are u a staff at moonwake now btw?
viral mason
Have they fixed the issue with the login thing where new users cannot use it or do you know
pliant stream
are u a staff at moonwake now btw?
Yes, you could say.
viral mason
What's moonwake?
pliant stream
Have they fixed the issue with the login thing where new users cannot use it or ...
Also i believe that has been fixed, the latest replay version doesn't have the login thing as far as i could tell
low shard
Alr lol, but any links referring to the site should be removed besides a downloa...
Weights stuff was already removed from the docs (except the security alert at risk), but most things should be removed here soon, i think just the model maker bot needs to remove it which I don't manage but it's not that important
low shard
Yes, you could say.
goodluck, how's it going? I remember checking it and it was less active to DreamTavern, I mean AI HUB isn't that active either
low shard
What's moonwake?
DreamTavern got rebranded
pliant stream
It's going alright, i was told not to talk about DreamTavern so i cant say much about it on that topic lol
low shard
It's going alright, i was told not to talk about DreamTavern so i cant say much ...
ohh alright, have a nice day
maybe let's not clutter the help channel, my bad for talking here
pliant stream
Yeah, fair enough lol.
viral mason
DreamTavern got rebranded
the ai image generator If I can remember right?
mild ridge
Is vonovox and wokada forks web ui only? or can it be run locally like the original wokada. i tried both and they launched the Web ui.
low shard
Is vonovox and wokada forks web ui only? or can it be run locally like the origi...
Original Wokad was also using Web UI, just in its own window
Vonovox doesn't use a Web UI, while Wokada Tg Develop does as its a wokada fork
Is it that important to you?
Btw its better you fill the help template form
mild ridge
- Goal : Educational purpose / Roleplay
- Specific Issue: Having have major audio issues and random high pitches, audio not being picked up properly (choppy) experience with vonovox.
- Full GPU Name:RTX 4070ti 12GB
- Operating System: Windows11
- Tutorial Link used: None / Previous knowledge.
cosmic ginkgo
for the voice models that are just a model, what am i supposed to input as the index for RVC?
fringe heron
for the voice models that are just a model, what am i supposed to input as the i...
leave it empty
cosmic ginkgo
will it work like that?
fringe heron
yes
cosmic ginkgo
oki ill give it a try ty :3
fringe heron
oki ill give it a try ty :3
np
viral mason
for the voice models that are just a model, what am i supposed to input as the i...
you could add the index of a different model, I dunno how that would effect it tho
cosmic ginkgo
what does the index effect?
fringe heron
you could add the index of a different model, I dunno how that would effect it t...
its not worth it
cosmic ginkgo
i usually have the index setting turned all the way down on every model i use anyway
i had another question, is there anyway to increase the quality of the voice that isnt the chunk?
fringe heron
i usually have the index setting turned all the way down on every model i use an...
lets say that when you make the model we make a map of how the target speaker speaks and we save it in a idnex file, that is used as a cheats sheet by the model to "remember how certein phonemes sounds"
cosmic ginkgo
upping the chunk makes the voice sound better but it takes soo long to translate
fringe heron
i had another question, is there anyway to increase the quality of the voice tha...
you can use extra convert size
viral mason
i usually have the index setting turned all the way down on every model i use an...
same here
viral mason
i had another question, is there anyway to increase the quality of the voice tha...
do u have AMD or Nvidia
if u have Nvidia u should use vonovox, it's the current best for realtime
fringe heron
upping the chunk makes the voice sound better but it takes soo long to translate
no, language is complex and takes a certain amount of time to recognize speech properly in the first place
cosmic ginkgo
lets say that when you make the model we make a map of how the target speaker sp...
oh okay i see i see, so theoretically like i could use my own voice if i knew how to make an indec and it would sound better when i use it?
i have nvidea
fringe heron
oh okay i see i see, so theoretically like i could use my own voice if i knew ho...
use your index with the model of someone else voice?
cosmic ginkgo
i think i use innxdirectml
fringe heron
no that wouldnt sound great
cosmic ginkgo
cuda
viral mason
where did you download the voice changer from btw
cosmic ginkgo
use your index with the model of someone else voice?
no like with my own, or am i not understanding it properly
fringe heron
no like with my own, or am i not understanding it properly
so the model trained to output your voice and the index of your voice?
cosmic ginkgo
you can use extra convert size
think i already use the max, then between 400-5000 chunk
*500
cosmic ginkgo
where did you download the voice changer from btw
i think from here? not sure bc i downloaded it a year or two ago and just started using it again
cosmic ginkgo
so the model trained to output your voice and the index of your voice?
yea i think so
viral mason
i think from here? not sure bc i downloaded it a year or two ago and just starte...
more than a year, probably either the original you're using or deiteris, I'll get u vonovox
cosmic ginkgo
why is vonovox better?
fringe heron
yea i think so
i mean yea but why would you want to convert you in you?
cosmic ginkgo
i mean yea but why would you want to convert you in you?
ohhh i see what u mean now
i think i get it
so i can just use the model and it will convert my voice anyway
like the index doesnt help much with that
fringe heron
so i can just use the model and it will convert my voice anyway
yes
fringe heron
like the index doesnt help much with that
the index for the most of the time is not worth using it
viral mason
why is vonovox better?
better quality, I believe less delay, it's the only one still receiving any updates or improvements, pitch detection and other stuff is way better too
cosmic ginkgo
the index for the most of the time is not worth using it
yeah on the UI i have it on 0 so
cosmic ginkgo
better quality, I believe less delay, it's the only one still receiving any upda...
oh okay cool!
fringe heron
better quality, I believe less delay, it's the only one still receiving any upda...
shame that it isnt open src
viral mason
shame that it isnt open src
true, but I believe it's not due to the patreon thing dr includes for optional support funding it
cosmic ginkgo
https://huggingface.co/dr87/vonovox/resolve/c8034f5f6d50648a8109bb4f847182362e2b...
do i just download these and launch them?
viral mason
do i just download these and launch them?
yea! the second link is a virtual audio cable, like vb cable, it's recommended over vb tho,
fringe heron
do i just download these and launch them?
yea, pretty much
viral mason
and for vonovox just run the start file
viral mason
wassup
cosmic ginkgo
yea! the second link is a virtual audio cable, like vb cable, it's recommended o...
OH okay i see, i just extracted it do i just launch the setup?
fringe heron
so since you have the model maker tag, i always made my models splitting dataset in clips of 3 sec to 5 sec, is it bullshit to split them in 400 ms or 300 to "optimize for small chuncks" when going realtime?
never really tried that
fringe heron
OH okay i see, i just extracted it do i just launch the setup?
yea
viral mason
OH okay i see, i just extracted it do i just launch the setup?
setup64, then install driver for vac lite
the virtual cable
cosmic ginkgo
setup64, then install driver for vac lite
64 over the regular setup?
viral mason
so since you have the model maker tag, i always made my models splitting dataset...
don't split up your dataset manually applio does it for you
viral mason
64 over the regular setup?
yea that's the correct file, idk why different ones exist
fringe heron
don't split up your dataset manually applio does it for you
i do know but i do it anyway because when loading datasets of 30+ hours ram cant take it
fringe heron
yea but in princible
never tried small chuncks?
always used the applio 1 sec min to 5?
cosmic ginkgo
is there anything i need to worry about like having installed or my pc specs or whatever before i launch vovonox? sorry im not that well versed v.v
fringe heron
is there anything i need to worry about like having installed or my pc specs or ...
you should be safe
cosmic ginkgo
i got a warning about not having smth installed in the cmd as it launched
fringe heron
post it
cosmic ginkgo
the cmd ent away like soon after, ill try launch again
Warning: psutil not available, cannot set CPU affinity
fringe heron
wait
cosmic ginkgo
oki
fringe heron
you runned setup.bat right?
cosmic ginkgo
WAIT
i might be just stupid
i was trying to use the launcher
not start.bat
LMAO
fringe heron
its alright
cosmic ginkgo
i feel like a hacker rn
fringe heron
you will need it
cosmic ginkgo
befre u go
fringe heron
yk?
cosmic ginkgo
can i pretty much just import all the models ove been using straight over to this?
fringe heron
yes
cosmic ginkgo
okay sweet
cosmic ginkgo
okay ty :3
well in that case
for the vac
does this mean its going to become my defulkt audio device ?
cosmic ginkgo
okok
fringe heron
so you need to go here, click on audio in the control panel
then simply select the audio device you had before or the one you always use and click set default wich for me its predefinito wich is italian
fringe heron
yes
cosmic ginkgo
will i need to do the same for my mic?
fringe heron
yes
cosmic ginkgo
okay sweet ty sm
fringe heron
np
fringe heron
hear the mdoel output?
viral mason
Sorry for disappearing btw I had to go to the store
cosmic ginkgo
its oki hahah
fringe heron
is there an option for audio playback on vovonox
just set the audio output to your headset
cosmic ginkgo
oh yea i see
viral mason
Then you cannot use it in games tho nobody but you could hear it
fringe heron
yea indeed, its playback
cosmic ginkgo
oh ok i see
fringe heron
just chnage it later to your virtual cable
cosmic ginkgo
so there is no way to do both? bc on my old one i was able to
hardy yew
just do this for your virtual audio cable
fringe heron
imput of cable to output and in game just output of virtual cable
hardy yew
and you will hear it
viral mason
just do this for your virtual audio cable
Go to control panel, audio settings, do this
fringe heron
just do this for your virtual audio cable
you can also do this yea
cosmic ginkgo
and then thje effects i shouldnt have to mess with right?
fringe heron
and then thje effects i shouldnt have to mess with right?
there is no need tbh
maybe
cosmic ginkgo
but the backend and sample rate etc is there any best option?
cosmic ginkgo
ok cool cool
viral mason
48k is fine, most default settings
Block size and pitch is all you'll need to change
fringe heron
sample rate use the model output or if you use the upscaler us 48khz
cosmic ginkgo
upscaler?
cosmic ginkgo
oh i see i see
fringe heron
(i raccomand not really using it)
fringe heron
the chunck
viral mason
Just a pitch shifter, only for fun
fringe heron
yes
viral mason
is there any dial for what it will pick up from my input?
Hm?
fringe heron
is there any dial for what it will pick up from my input?
what
cosmic ginkgo
Just a pitch shifter, only for fun
oh okaty, assu,med that was just thje p[itch
viral mason
Pitch is for making it sound correct for your voice, formant is a pitch shifter
viral mason
Male to female 3-12 female to male same but negative
Male to male 0
Same female
Depends on if you're a guy or girl and the model being female or male
That's for pitch
cosmic ginkgo
okay sweet
viral mason
Block size (chunk) should be fine at 0.30 or 0.35
cosmic ginkgo
is higher better qaulity? i assume
fringe heron
ye
cosmic ginkgo
sweeet
fringe heron
till a point tho
cosmic ginkgo
yea
fringe heron
for me some models higher would make it sound a bit odd
cosmic ginkgo
oh really?
fringe heron
but for me its always a pain to get them relaible on my voice
viral mason
is higher better qaulity? i assume
Depends, from testing the models get weird after maybe 0.50
fringe heron
Depends, from testing the models get weird after maybe 0.50
@viral mason , all the models you did that you tested on your voice did they sounded actually reliable or they would produce garbled on some imputs?
or perceptually bad audio
viral mason
<@1023278814752677918> , all the models you did that you tested on your voice di...
Yea they're completely reliable because I made them lol
fringe heron
Yea they're completely reliable because I made them lol
🤨
viral mason
Do you have a noisy environment or people talking/tv in the background
fringe heron
nope
viral mason
Anything that could cause it
cosmic ginkgo
sometimes mine picks up my keyboard
viral mason
Same
viral mason
most likely is my voice the problem
What voice model are you using?
fringe heron
my physical voice
cosmic ginkgo
doesnt help i SMACK my keys ig
fringe heron
at infer as source
viral mason
You made a voice model of yourself?
fringe heron
no
what i mean
is that any vocie model if i convert my vocie with it and i speak naturally it always ends up on some wierd soudns sometimes
unless the target is very similar to me
fringe heron
btw i am not a native english speaker and my english could be a bit broken) but using same acent speaker targets and gettign same results is ass
cosmic ginkgo
i know mine does that
i have an australian accent so
some words sound really odd
cosmic ginkgo
and also if i laugh it alwasy sounds like deoms
fringe heron
and i might have found it
cosmic ginkgo
rlly?
fringe heron
and also if i laugh it alwasy sounds like deoms
i mean non verbal are always a pain
cosmic ginkgo
yea haha
cosmic ginkgo
yea but its not trivial
i cant imagine it would be
fringe heron
and requires quite some work
viral mason
and also if i laugh it alwasy sounds like deoms
Laughs are not possible really unless a model is trained on it
cosmic ginkgo
my best bet has just been to adjust my voice when i use the rvc sand have my mic closer
fringe heron
i mean in theory its easy but requires a lot of work and i dont knwo hwo relaible it is right now
cosmic ginkgo
Laughs are not possible really unless a model is trained on it
yeah makes alot of sense
fringe heron
Laughs are not possible really unless a model is trained on it
kinda
fringe heron
If it's not in the dataset it'll usually sound off or just not natural
with my trainign method the dataset can cointain very few of them and maybe sound fine
viral mason
Yea as long as it's in there should be fine
fringe heron
but for me (to my current looks i gave) the problem is that finetuning a pretrained shits what makes the model "speaker invariant"
viral mason
?
viral mason
you know how the model works?
Not really, I'm not into code I just know how to make a good model
fringe heron
oh well
fringe heron
perhaps is not the te but contentvec
the TE is what paints how a phone and pitch would bee if spoke by someone
viral mason
I hope real spin is implemented to applio soon
fringe heron
the TE is what paints how a phone and pitch would bee if spoke by someone
its like taking a pic of a car and slapping a red filter on it or a blue one
even if contentvec is not perfect
the content is pretty much the same
and if you finetune the TE on the target then when sum else uses it that person is not the target
if it goes out of distribution the decoder struggles or even if the output latent is wierd because its not sure how it would be
i will test this better soon
fringe heron
I hope real spin is implemented to applio soon
fr
proven hill
I hope real spin is implemented to applio soon
is that much of a game changer?
i always trained “vanilla” and got similiar result to “improved” versions
cosmic ginkgo
@fringe heron @viral mason ty sm for helping me out im having so much fun playing around with it :3333
viral mason
is that much of a game changer?
The real spin won't have the timbre leak and other weird issues, lyere could explain
fringe heron
<@1361372948463878367> <@1023278814752677918> ty sm for helping me out im having...
:)
proven hill
The real spin won't have the timbre leak and other weird issues, lyere could exp...
understood
analog obsidian
i always trained “vanilla” and got similiar result to “improved” versions
applio's spin is not the same as the original spin
proven hill
applio's spin is not the same as the original spin
why?
proven hill
different projection and doesnt use pred head
gotcha
analog obsidian
so it behaves like a worse contentvec lol
ok to be fair, not worse, but the quality of the feature extraction is not as good as the original spin
proven hill
so it behaves like a worse contentvec lol
ok to be fair, not worse, but the qual...
mhhh thats why my result was bad
fringe heron
btw i also realized this, you said you tried freezing the TextEnc but didnt notice improvements, maybe you kept the posterior enc free to update wich maybe caused a easy path to be "speaker variant" again and saw no improvments. Even if thinking more about it would only cause more problems freezing it.
raven bluff
where do i ghet the voicechanger
opaque pond
Please can I get a link to set up vonovox…. On a nvidia 4060, 8gb vgram…. What’s the best set up for real time voice changer?
viral mason
Please can I get a link to set up vonovox…. On a nvidia 4060, 8gb vgram…. What’s...
sure, what are u planning on using it for btw
low shard
Please can I get a link to set up vonovox…. On a nvidia 4060, 8gb vgram…. What’s...
!help-template
patent trellisBOT
!help-template
low shard
Please can I get a link to set up vonovox…. On a nvidia 4060, 8gb vgram…. What’s...
!help-template
patent trellisBOT
!help-template
viral mason
we don't need 3
hallow thistle
where do i ghet the voicechanger
What is your PC GPU? And what do you use the voice changer for?
vale harbor
Is Vonovox having issues? It lags when I play games, and I'm using a 16 GB RX 5070 Ti.
My internet connection is pretty good; the lag only goes away when I close Vonovox.
undone dune
- Goal (e.g., TTS, AI Covers, Roleplay): just wanna sound like anime characters
- Specific Issue: dont know what to download i wanna download voicechanger
- Full GPU Name: Nvidia Gforce rtx 5070
- Operating System: win 11
- Tutorial Link used: https://www.youtube.com/watch?v=81KYc8AAmus
hallow thistle
- Goal (e.g., TTS, AI Covers, Roleplay): just wanna sound like anime characters ...
Vonovox and Tg Develop's voice changer are only known voice changers that can work with GeForce RTX 50 series. 
undone dune
Vonovox and Tg Develop's voice changer are only known voice changers that can wo...
could u send me the link please
hallow thistle
Download Vonovox. https://huggingface.co/dr87/vonovox/resolve/main/Vonovox_beta_17_11.zip
vale harbor
Vonovox and Tg Develop's voice changer are only known voice changers that can wo...
So why is it lagging?
hallow thistle
So why is it lagging?
Can you send the full screenshot?
ebon basin
any1 got a mobile voice changer that’s not bad
hallow thistle
lilac moss
- Goal (e.g., TTS, AI Covers, Roleplay): Extended Pretraining of base rvc v2
- Specific Issue: Not really sure how to do this, any guidance? Do I just plug in the D and G models into applio? A CLI option for this would be great..
- Full GPU Name: A100 80GB
- Operating System: Linux (specifically debian)
- Tutorial Link used: N/A
misty trellis
Goal, get good tts be it local or using providers like eleven lab
Silly tavern and marinara engine has little to nothing guided and etc on tts
Gpu is rx6900xt 16gb vram
Os is windows 11
None
low shard
@shut yoke don’t promote
misty trellis
<@1500051865176637513> don’t promote
Wait huh did something happen
low shard
any1 got a mobile voice changer that’s not bad
There's no realtime voice changer for mobile
misty trellis
Oh my bad I saw original message deleted so I thought that was it 🤣
low shard
- Goal (e.g., TTS, AI Covers, Roleplay): Extended Pretraining of base rvc v2
- S...
It's more of an advanced thing but you could check https://docs.aihub.gg
low shard
So why is it lagging?
!help-template
patent trellisBOT
!help-template
lilac moss
It's more of an advanced thing but you could check https://docs.aihub.gg
I don't need to know it's advanced lol I checked the docs, nothing much was said there. Just the same advice,, use multi-speaker on applio, bla bla bla, do I just plug in the generator and discriminator from v2 and just chug on?
simple ore
I don't need to know it's advanced lol I checked the docs, nothing much was said...
if you use D/G weight to train a model, it is called a finetuning
you can finetune using a single voice.. or using 100+
if you dont use D/G weights and train using 100+ voices, it is called 'creating a pretrain'... nobody has been successful at creating a good one from scratch yet
lilac moss
if you dont use D/G weights and train using 100+ voices, it is called 'creating ...
.. How come all these pretrains like Titan etc crop up then? Those are trained ON rvc v2 aren't they? That is what continued pretraining is no?
misty trellis
Goal, get good tts be it local or using providers like eleven lab
Silly tavern and marinara engine has little to nothing guided and etc on tts
Gpu is rx6900xt 16gb vram
Os is windows 11
None
simple ore
.. How come all these pretrains like Titan etc crop up then? Those are trained O...
they are trained on top of OG pretrain... usually without the same variety, so they kinda suck
lilac moss
if you use D/G weight to train a model, it is called a finetuning
iirc finetuning is single-speaker isn't it? + it doesn't necessarily "teach" phonemes or language or necessarily new data towards the model
lilac moss
they are trained on top of OG pretrain... usually without the same variety, so t...
hmm ive heard
good words abt these pretrains
still, could you lmk how to do it?
docs aren't very helpful this time
simple ore
do you have 100+ different voices with a wide range of phonemes in their respective datasets?
with equally good audio quality?
if yes, then you can re-tune the og pretrain
lilac moss
yes
lilac moss
do you have 100+ different voices with a wide range of phonemes in their respect...
i have around 150 - rvc v2 was trained with around 108 or so? does applio cover the extension of the tensor in this case?
hallow thistle
Challenge accepted?
simple ore
i have around 150 - rvc v2 was trained with around 108 or so? does applio cover ...
like this
and use this new g + original d
simple ore
you need to run this script and match the number of speakers in the prepared dataset
Applio can only adjust when you try from scratch
lilac moss
you need to run this script and match the number of speakers in the prepared dat...
kay, then do i just plug it into applio and train on the g and d?
simple ore
you use the extended g and original d as a custom pretrain
if you're training locally, you can simply run a .bat file
env\python rvc\train\train.py VCTK_32k_SP1024 1 20 rvc\models\pretraineds\hifi-gan\f0G32k_emb129.pth rvc\models\pretraineds\hifi-gan\f0D32k.pth 0 16 32000 False True False False 5 False "HiFi-GAN" False
like this
lilac moss
`env\python rvc\train\train.py VCTK_32k_SP1024 1 20 rvc\models\pretraineds\hifi-...
I'm assuming I replace the obvious values like 32k and all with my actual sample rate?
simple ore
yes
simple ore
model_name = sys.argv[1]
save_every_epoch = int(sys.argv[2])
total_epoch = int(sys.argv[3])
pretrainG = sys.argv[4]
pretrainD = sys.argv[5]
gpus = sys.argv[6]
batch_size = int(sys.argv[7])
sample_rate = int(sys.argv[8])
save_only_latest = strtobool(sys.argv[9])
save_every_weights = strtobool(sys.argv[10])
cache_data_in_gpu = strtobool(sys.argv[11])
overtraining_detector = strtobool(sys.argv[12])
overtraining_threshold = int(sys.argv[13])
cleanup = strtobool(sys.argv[14])
vocoder = sys.argv[15]
checkpointing = strtobool(sys.argv[16])
tysm! i was genuinely writing my own rvc train implementation and it was hurting my head so bad haha 😭😭
short panther
Do you have huge differences in AI use in your teams? Some are eager to learn new things, some are still about to explore ChatGPT... how to bridge the gap?
low shard
you need to run this script and match the number of speakers in the prepared dat...
so they would need to adjust the script to finetune correctly with expansion on rvc v2? Would that get added in the next applio version?
simple ore
normal users do not finetune datasets with 100+ speakers
if you use less than 110 you'll be fine 🙂
low shard
normal users do not finetune datasets with 100+ speakers
do you think that script might be needed to get added in the ai hub docs? It’s a pretty rare case
simple root
Can any 1 guide me throughly on using Flow
lilac moss
do you think that script might be needed to get added in the ai hub docs? It’s a...
please do honestly, ai hub docs are really sparse when it comes to niche stuff like cpt
simple ore
do you think that script might be needed to get added in the ai hub docs? It’s a...
very unlikely
but if you want to, go ahead
hallow thistle
I just wanted to share the thing. There's a strategy where if your laptop doesn't have a dedicated GPU, you'd go for an online service like Kaggle or Google Colab, so it should make the voice changer to work perfectly in theory. Actually, when I do the same as this current 2012 Dell laptop that has second gen Intel Core i3 CPU, the audio still stutters because CPU has to process audio stream and other programs at the time, usually unbearable in actual runs.
misty trellis
Goal, get good tts be it local or using providers like eleven lab, fish audio for example
to get Silly tavern and marinara engine with working tts
Gpu is rx6900xt 16gb vram
Os is windows 11
None
hallow thistle
So, if I actually have to run the voice changer, at this point I'd buy a new PC with a GPU instead. 
I don't feel like I need help about this one, I already know how things work, even if I rarely run the voice changer for anything other than as a dummy program.
signal jetty
Hello can I post a repo to help me?
opaque pond
I did everything I saw online right but I’m still getting bad lag issues
opaque pond
sure, what are u planning on using it for btw
Obs for streaming
hallow thistle
Hello can I post a repo to help me?
What is this repository?
karmic trellis
Hi. I have been using a voice model for a while now (with vonovox beta, without any index file, because there was no index file attached to the voice model when i downloaded it), but its quality is sometimes not the best, is there anything i could do to make it better? Like training the voice more myself, enhancing it in any way? It sounds robotic sometimes. I am absolutely willing to spend time on making it better or to learn new things i dont know how to do now, I just want to ehance it somehow
hallow thistle
Why roleplay?
hallow thistle
Did you follow a guide or tutorial before?
viral mason
-rt
lusty bear
hello, how can i fix my voice chopping of when using a model, its sounds good but every half a second it bugs, i have an rx6600 and a nice maono mic so idk whats causing this issue
viral mason
hello, how can i fix my voice chopping of when using a model, its sounds good bu...
What voice changer are u using rn? You could have an older one that doesn't run as good as some of the new stuff
lime nova
hello, how can i fix my voice chopping of when using a model, its sounds good bu...
If it's for sure the model and not your mic being too hot or OBS then most likely is the block size being too small If the GPU can't finish processing a block before the next one arrives you get dropouts
viral mason
How do people literally disappear after asking for help, they must not be in a hurry
fringe heron
hello, how can i fix my voice chopping of when using a model, its sounds good bu...
what kind of "bugs" you hear?
low shard
hello, how can i fix my voice chopping of when using a model, its sounds good bu...
!help-template
patent trellisBOT
!help-template
lusty bear
How do people literally disappear after asking for help, they must not be in a h...
I had the best beans of my life, i guess its kinda annoying tho
viral mason
W beans
viral mason
that one is super outdated, what gpu do u have? (Nvidia or AMD)?
viral mason
k one sec I'll get u the downloads to the newest thing
viral mason
yup
first one is the voice changer, second one is a virtual cable that connects it to games or discord ect
lusty bear
is VBcable bad?
viral mason
I actually have to go somewhere rn so I will not be avaialble consistently, for quick setup run setup64 for vac lite and for tg fork run mmvcserversio
viral mason
is VBcable bad?
not really but it can be weird at times and cause odd mic issues
viral mason
imma follow you up when i set it up
back home, any problems or anything?
proven hill
back home, any problems or anything?
yes i have a problem
viral mason
let me hear it :3
proven hill
let me hear it :3
bababui
lusty bear
back home, any problems or anything?
im also out rn, so i havent tested anything out
viral mason
did you close the command promt?
lusty bear
no
viral mason
hmm
what are you microphone settings like?
input should be your regular mic and output should be line 1
viral mason
do u have a model installed?
viral mason
client doesn't work for me at all but try both
lusty bear
still getting the same pipeline error
on both
[VoiceChangerManager] 'Pipeline is not initialized.'
fringe crag
hello
if i use general questions, in an offline model, how to know that i have the right model for my build? Is there like a sort of response time? For example, asking how much 1+1 is, and i get an answer in 1 sec or in 1 minute ... /: How can i know this?
viral mason
[VoiceChangerManager] 'Pipeline is not initialized.'
I'm not sure how to fix errors, <@&1159293204038955078> help this man
lusty bear
I'm not sure how to fix errors, <@&1159293204038955078> help this man
yea, what he said
outer sparrow
Hey plz where is the site for downlaod vc
low shard
if i use general questions, in an offline model, how to know that i have the rig...
do you want to know the right model for your task or the performance for your hardware?
low shard
Hey plz where is the site for downlaod vc
this is a General AI Discord Server, there isn't a single voice changer program
!help-template
patent trellisBOT
!help-template
lusty bear
[VoiceChangerManager] 'Pipeline is not initialized.'
@low shard could you help me on this?
outer sparrow
this is a General AI Discord Server, there isn't a single voice changer program
thanks
fringe crag
do you want to know the right model for your task or the performance for your ha...
i guess, a bit of both... i want mainly working with texts (summarising, explaining books, ...) So, while i think in LLM standards, thats not of the heaviest processes, i also don't want to choose a model where i have to wait 2 minutes on my answer
lusty bear
https://docs.aihub.gg/realtime-voice-changer/local/tg-develops-w-okada-fork/#pip...
i download this?
low shard
i download this?
it's not something to download, it's usual fixes to your issue
low shard
i guess, a bit of both... i want mainly working with texts (summarising, explain...
you could check posts about the models, https://lmarena.ai, and https://www.localllm.run/
fringe crag
ok, thanks
misty trellis
Goal, get good tts be it local or using providers like eleven lab, fish audio for example
to get Silly tavern and marinara engine with working tts
Gpu is rx6900xt 16gb vram
Os is windows 11
None
shy mica
proven hill
Goal, get good tts be it local or using providers like eleven lab, fish audio fo...
it wouldve been easier with a nvidia card
misty trellis
it wouldve been easier with a nvidia card
i do have but it's a latop 3050...
proven hill
i do have but it's a latop 3050...
:/
lusty bear
I'm not sure how to fix errors, <@&1159293204038955078> help this man
is there another voice changer i could use?
maybe this one doesnt like me lol
viral mason
is there another voice changer i could use?
Nothing good tbh, you could try Wokada deiteris
It's not as good as tg fork tho
low shard
it's not something to download, it's usual fixes to your issue
@lusty bear did u check this?
molten spire
Is this the right channel to ask for pre train models recomendations?
lusty bear
<@1007023637347450940> did u check this?
yes
i had some
updates
the model is selected
doesnt have any special characters
low shard
updates
did you update everything, select an RVC Model before starting the server, and are using no weird character?
lusty bear
did you update everything, select an RVC Model before starting the server, and a...
is kanye weird?
low shard
could you send a screen recording of you starting the program?
lusty bear
like when i get the error or?
low shard
is kanye weird?
I meant special characters like '£'
lusty bear
no
low shard
like when i get the error or?
From when you open the .exe to when you start the server
lusty bear
this is the error
low shard
have you checked if the same issue appears on every model?
If so, have you tried re-downloading wokada tg develop?
lusty bear
oh, one thing
the sample rate auto changes to 44100 when i start the server even when i set it to 48000
low shard
the sample rate auto changes to 44100 when i start the server even when i set i...
it might be a corrupted file in the essential files downloaded on first startup, could you try re-downloading the program ?
lusty bear
it might be a corrupted file in the essential files downloaded on first startup,...
sure
this one right?
vale harbor
Is this lag normal when I open a game? It didn't used to happen to me before.
Full GPU Name: rtx 5070 ti 16 gb
- Operating System: windows 11 pro
viral mason
Full GPU Name: rtx 5070 ti 16 gb
- Operating System: windows 11 pro
I use the same setup, mine doesn't lag
hi again
have you tried turning the graphics of the game down?
vale harbor
have you tried turning the graphics of the game down?
Lower the graphics settings?
vale harbor
I use the same setup, mine doesn't lag
Or is it because I have the virtual audio and not the other one they sent me?
viral mason
I don't think the cable is the issue
try turning down the graphics in game to see if it helps preformance
vale harbor
try turning down the graphics in game to see if it helps preformance
I also set this up
viral mason
you don't need these tbh
viral mason
this one is fine at default
vale harbor
this one is fine at default
how is default?
vale harbor
you don't need these tbh
why
viral mason
why
they're filters all they do is change how it can sound, no quality improvements or anything
vale harbor
they're filters all they do is change how it can sound, no quality improvements ...
And aren't they meant to make you sound more like the character?
viral mason
no they're more just to help with background noise or for the eq it modifies sound by increasing or decreasing the loudness of specific pitches to improve clarity, or alter the tone to be warmer, brighter, or less harsh.
that last part I took from google
vale harbor
no they're more just to help with background noise or for the eq it modifies sou...
So, if that helps, I have some background noise and, among other things...
vale harbor
this one is fine at default
So it doesn't require overclocking the CPU?
viral mason
why would you do that?
bronze pier
are there any google colabs for training voice models that still work? or is Applio the only way
viral mason
are there any google colabs for training voice models that still work? or is App...
applio is the best way to train, colab is ok but Kaggle is way better
graceful scaffold
hi guyssss
viral mason
heloo
graceful scaffold
whatcha talkin about/what can i help with :3
woah, ai master :O
have you mastered the arts of the shoggoth?
viral mason
hmmm? what's that? :3
proud vapor
One message removed from a suspended account.
One message removed from a suspended account.
hallow thistle
whatcha talkin about/what can i help with :3
graceful scaffold
anyway sad to see weights shut down last month, i recorded the final minutes
viral mason
at least replay is fully functional without needing an account now, only good thing weights ever made was that
hallow thistle
So it doesn't require overclocking the CPU?
Not even needed.
vale harbor
Not even needed.
hallow thistle
hallow thistle
anyway sad to see weights shut down last month, i recorded the final minutes
"Weights.com closure" might sound too overdramatic for some, but in the end it was just another website, so either try move on or ask for AI cover program. 
hallow thistle
Are you sure those options aren't needed?
Not needed.
vale harbor
Not needed.
graceful scaffold
question for yall
whats the max time on a google colab instance or whatever?
is it till i close the tab orrrr
viral mason
like 4 hours I think
kaggle gives 30 hours per week for free users ❤️
much better
and easier to monitor
hallow thistle
whats the max time on a google colab instance or whatever?
4 hours average in a day.
silent grove
It sucks that I have all these rvc voices saved but no program to make covers with them
All around just sucks
hallow thistle
There's Applio RVC.
hallow thistle
viral mason
It sucks that I have all these rvc voices saved but no program to make covers wi...
have you heard of replay?
viral mason
it was made by weights actually and works the same
silent grove
YOU FOOL
viral mason
u go to weights and the download is at the bottom
hallow thistle
viral mason
hallow thistle
For real though, Applio RVC can be a bit harder to use than Replay, but Applio RVC can also be used to train a voice model in one instance.
viral mason
very true
if he is just looking for a quick cover application replay is nice and requires very little skill to use
graceful scaffold
if he is just looking for a quick cover application replay is nice and requires ...
fun fact, i just downloaded replay to try it out
and see if its any better than aicovergen or an RVC webui
graceful scaffold
cant post images :<
nvm
THIS FAT PROGRAM LOL
btw guys i recommend "Mem Reduct" its pretty good and can free up RAM
hallow thistle
btw guys i recommend "Mem Reduct" its pretty good and can free up RAM
I'm quite skeptical with this "free RAM" recommendation. Sure, the Mem Reduct might work for a low-end PC (if has 8 GB) as it claims, but for a PC with around 12GB or more I don't think there's a point to have it.
You still haven't told about your laptop specs.
simple raptor
- Goal (e.g., TTS, AI Covers, Roleplay): LoRA training - Generation of reference images
- Specific Issue: Can't get QwenVL custom node working
- Full GPU Name: NVidia GeForce RTX 4080
- Operating System: Windows 11
- Tutorial Link used: https://www.youtube.com/watch?v=WRaOsu9TDEM&t=160s
I have tried for days to setup/configure my comfyui environment to be able to run QwenVL custom node and models. Running the workflow in the tutorial (approx. 20:34 in timeline) I get missing packages (accelerate) and incompatible package versions (torch) as well as not being able to find the correct GGUF models. I would appreciate assistance in getting this sorted. I've been using CoPilot and 'he' has been doing my head in going overboard with possibilities BUT he has identified that my biggest issue is package versions given I am trying to keep my environments as clean as possible. Can someone provide a 'pip freeze' of their environment, that would go a long way to helping? Thanks Steve
PS - can anyone identify the 3 x '..._Sub' nodes used in the ZIT, Qwen2512 and Flux Klein groups?
graceful scaffold
You still haven't told about your laptop specs.
i use it quite often (mem reduct) actually, 32gb ram and 6gb vram on a laptop nvidia geforce rtx 4050
viral mason
never heard of it
graceful scaffold
never heard of it
guys theres a worm outside my house
viral mason
:3
graceful scaffold
ima go 2 bed
hallow thistle
i use it quite often (mem reduct) actually, 32gb ram and 6gb vram on a laptop nv...
The Mem Reduct isn't really needed for this system of 32 GB anyway. Your "dedicated" NVIDIA GPU is decent enough for simple AI cover, although not preferred to train anything. I've never used Replay, but if Replay couldn't detect your laptop GPU, there's Applio RVC as another option.
patent trellisBOT
viral mason
if you're looking for the voice changer use vonovox for Nvidia gpus but for AMD u can't use Vonovox so use Wokada tg fork
viral mason
what is ur gpu?
buoyant thistle
i just deleted the original wokada cuz it says its not good
viral mason
i just deleted the original wokada cuz it says its not good
it's not lol, super unoptimized and stuff
buoyant thistle
ya just some fun stuff, nothing professional
viral mason
thank god
so many weirdos join here to play as "girls"
everyone knows that means they wanna scam ppl
buoyant thistle
they want the method
viral mason
and we don't allow that here
buoyant thistle
oki so u recommend nvidia tg fork
buoyant thistle
i am upgrading to a 5070 ti and ryzen 7 9800x3d soon tho, im assuming i wouldnt use that nvidia tg fork?
bronze pier
for Applio online tensorboard, what do i put for dataset path? i'm running on the public url since the private one doesnt work
viral mason
just use the gpu instead on it?
gpu is the main thing tbh
viral mason
for Applio online tensorboard, what do i put for dataset path? i'm running on th...
what are u running it on?
colab? locally on pc? Kaggle?
bronze pier
I downloaded Applio and uploaded colab to my google drive and ran it there
viral mason
I am confused
buoyant thistle
gpu is the main thing tbh
do u have the link for the one we were talking abt
bronze pier
im running this file on google colab
its in the Assets folder when i downloaded Applio
viral mason
nvidia tg fork
extract the first one then place the second one in the folder for the first one, third link is a virtual audio cable that is used like VB cable to connect the voice changer to discord or games ect
buoyant thistle
if u have trouble or questions just ping me
fanks
devout dome
hey i keep seeing people talk about "crossfade" and "extra time" but i only see block size. is there any reason why?
- Goal (e.g., TTS, AI Covers, Roleplay): Roleplay
- Specific Issue: i keep seeing people talk about "crossfade" and "extra time" but i only see block size. is there any reason why?
- Full GPU Name: RTX 5070 12GB
- Operating System: Win 11
- Tutorial Link used: None
viral mason
hey i keep seeing people talk about "crossfade" and "extra time" but i only see ...
That's because you're most likely using the Vonovox beta, which removes those and have them set at the best option already internally
So those don't matter
devout dome
ahh i see
oh about block size is there like an area i should keep it at for it to not have too big a delay while still sounding ok
im using your teto model atm
devout dome
That's because you're most likely using the Vonovox beta, which removes those an...
sorry i forgot to hit the reply button
viral mason
oh about block size is there like an area i should keep it at for it to not have...
U can try around 0.30 to 0.50
More or less than that it kinda starts sounding bad
0.25 is also ok
I personally find 0.35 good as well
devout dome
is this the same way while gaming?
viral mason
Ya
viral mason
Nope lol, gotta yap to level up
You're very welcome! If u have more questions or need help or anything u can message me here :3
devout dome
You're very welcome! If u have more questions or need help or anything u can mes...
will do and ty again for the help and the model it sounds really well :D
devout dome
@viral mason
sorry for bothering but I was looking at your GLaDOS model and I saw a fx file for fl studio. is there a way for me to get effects to work on vonovox so i could use it in real time or is that not possible rn?
low shard
this one right?
How's it going?
hallow thistle
and im a girl now
low shard
and im a girl now
Are you trans or looking to troll?
hardy yew
<@1023278814752677918>
sorry for bothering but I was looking at your GLaDOS mod...
What Worm does (if I remember correctly) is route vonovox's output to FL studio and then use output from FL as input in e.g. games
Anyway, definitely possible
Not sure what he does (he's probably sleeping now and will respond later) but I guess 2 virtual audio cables would do the job
Unless FL studio ships with something like that built-in
Definitely higher delay than without FL in between
But opens up room for a ton of effects
And glados sounds great with that autotune
viral mason
<@1023278814752677918>
sorry for bothering but I was looking at your GLaDOS mod...
The effect used for her specifically costs money I'm sorry, also you'll need Voicemod, Fl studio, Vonovox/Wokada, vac lite, VB cable, and voicemeeter
I can't help showing the setup atm as I'm already in bed and turned my computer off
It also requires VB cable doesn't kill itself upon use
Tried the same setup I have on my pc on my laptop and the one on laptop explodes and fizzles out
VB cable is strange
hardy yew
Ooh not asleep after all!
hardy yew
If VB cable is a pain then paid VAC offers up to 256 cables so xD might be worth it for someone that needs it often
It's not expensive IIRC
viral mason
Icky
No money for my setup that's bad for business
It must stay free for the people
The uh, two paid voice effects don't count btw uhmm 👀
They're for personal use on models
hardy yew
The uh, two paid voice effects don't count btw uhmm 👀
I was about to mention them
Too late
xD
viral mason
Hehe beat you to it
But ye the autotune and ultrapitch are basically to make the voices sound more accurate
Autotune for Glados and Cyn and ultrapitch for anything droid related
hardy yew
Yee I remember the glados example was awesome
viral mason
:3
I really should remake some of those droids
They're funny
The b1 is hell to go through over an hour of audio from just 1 game
I want more of General Grievous in games but nobody ever focuses on clone wars era anymore
viral mason
Yee I remember the glados example was awesome
The turrets also are quite silly with autotune, I've recently made the effect more tame over time since it was too much before, u can see the change over time especially for glados
low shard
i reinstalled the program and now it works
you're welcome, seems like some of the files didn't download correctly at first lol
lusty bear
you're welcome, seems like some of the files didn't download correctly at first ...
yea
lavish python
my gpu is 5070 12gb, will using average models increase much input latency on games like valorant
pulsar token
my gpu is 5070 12gb, will using average models increase much input latency on g...
not really if u lock ur fps
lavish python
not really if u lock ur fps
but it will increase input latency right?
is it that much latency to consider?
pulsar token
a little bit yeah, but I doubt it will be noticeable if u put fps limit
lavish python
a little bit yeah, but I doubt it will be noticeable if u put fps limit
how about gpu usage? will it use as much as gpu resource available?
on average models
hallow thistle
my gpu is 5070 12gb, will using average models increase much input latency on g...
What model?
lavish python
What model?
this one for example
the game consumes only 20-30% gpu usage
is input latency affected if i use in game?
hallow thistle
Applio RVC, Vonovox or W-Okada voice changer?
lavish python
wokada
hallow thistle
Which version is your W-Okada?
lavish python
v.2.2.2-beta
viral mason
Old and outdated as hell
lavish python
for example
hallow thistle
Vonovox and Tg Develop's voice changer (b2397) are only known voice changers that can work with NVIDIA GeForce RTX 50 series so.
viral mason
for example
U should switch to Vonovox, with your gpu it'll run much better and it's the current best real-time voice changer
lavish python
oh alr
lavish python
https://discord.com/channels/1159260121998827560/1498564814149128393
so this one
i need an answer for this
hallow thistle
That's basically another RVC voice model. Almost every RVC model works pretty much the same. The audio latency depends on not just the settings.
lavish python
Here's the two downloads you'll need
https://huggingface.co/dr87/vonovox/resolve...
thanks
yeah i gonna research it deeper later
but before that
i have to know
using voice changer will increase input latency or not?
viral mason
Like Namari said it's just another voice model, doesn't affect how the voice changer would work
lavish python
i dont matter audio latency
viral mason
using voice changer will increase input latency or not?
That's not how they're supposed to
lavish python
input latency from mouse i mean
hallow thistle
lavish python
i thought when i use high graphics push gpu usage further make higher latency
so using voice changer would be the same way
voice changer also push gpu usage further
viral mason
Don't use high graphics when using a voice changer and gaming
It's not good for either program
hardy yew
using voice changer will increase input latency or not?
Nah, you're fine as long as you're leaving reasonable headroom and not bottlenecking
lavish python
valorant usage on gpu is only 20-30%
so this is totally fine to use voice changer beside
is this right?
hardy yew
Most likely yeah, if your usage is up to 30% then you definitely already limit framerate
Should be good
hardy yew
But anyway, just run the voicechanger and if you still have a bit headroom it's all good
If not, lower the settings or further limit the framerate
Anything that will lower the load where it struggles
lavish python
oh thank yall for answer
i found sometime the gpu jump up to 50% in combat
anyway to limit gpu usage of voice changer?
so i wont get laggy on combat
hallow thistle
Make sure to try Vonovox.
devout dome
I can't help showing the setup atm as I'm already in bed and turned my computer ...
ah i see, tho, if possible whenever you get the chance could you show the setup or explain it in more detail if i have the chance or time to do it?
lavish python
Make sure to try Vonovox.
bet i will try
bronze pier
Whats the easiest way to train a voice model besides Applio?
viral mason
viral mason
ah i see, tho, if possible whenever you get the chance could you show the setup ...
Sure I could show it at some point, the whole thing is just kinda a lot of vac and VB cable connection to many things but it does cause latency increases
stable basalt
What is the best free rvc to use rn
devout dome
Sure I could show it at some point, the whole thing is just kinda a lot of vac a...
nice ty so much :D
hallow thistle
Whats the easiest way to train a voice model besides Applio?
No idea.
viral mason
What is the best free rvc to use rn
Wdym? For making rvc models or for realtime use as a voice changer?
Also what gpu do you have (Nvidia or AMD)
And what are u wanting to use it for?
stable basalt
Nividia
stable basalt
Wdym? For making rvc models or for realtime use as a voice changer?
Real time
stable basalt
And what are u wanting to use it for?
Trolling
viral mason
Aw :c
stable basalt
With friends
stable basalt
What
viral mason
Ye I saw this art on insta and thought it looked very beautiful
The full image is very nice
stable basalt
What is the best free rvc to use rn
viral mason
What is the best free rvc to use rn
You should elaborate on the previous statement you made
stable basalt
What statement
devout dome
The full image is very nice
sick i dont see many BBQ artist on my insta
for me its usually vocaloid stuff
viral mason
What statement
Why would you want to use it for trolling, what even is it for
stable basalt
Why would you want to use it for trolling, what even is it for
Why are you asking?
Is this any of your business?
viral mason
sick i dont see many BBQ artist on my insta
If u like a few Ena posts you'll get flooded with it eventually
viral mason
Why are you asking?
Because if you read the rules you might know that if you're planning on using those e-girl models for such is not allowed here
devout dome
Is this any of your business?
maybe just to see if ur gonna be doin too much (bad stuff) with the vc
stable basalt
I am not
viral mason
Then explain
stable basalt
Was gonna use markiplier voice
hallow thistle
You could say that.
viral mason
I'm not a mod sadly so I cannot help with that
But the best Nvidia voice changer to use for free is Vonovox
devout dome
Because if you read the rules you might know that if you're planning on using th...
why are the egirl models posted here instead of a different area it seems counter productive no?
stable basalt
Got download link?
viral mason
why are the egirl models posted here instead of a different area it seems counte...
It's not technically breaking any of the servers rules to make them sadly
So we just go like eww to whoever makes the model and move on
If I was a mod I'd delete the links to all e-girl models immediately
If the weirdos need them so bad go find them yourself
viral mason
Got download link?
It should be posted above, just scroll a bit to find it
devout dome
If I was a mod I'd delete the links to all e-girl models immediately
now where would teto, miku, and vtubers + every other (technical e-girl) go? wouldnt they be included?
i say technical e-girl because arent there some that are quite literally just a female voice
viral mason
That's different
devout dome
like i think callie? idk
viral mason
They are actually cool unlike "whispery mommy girl voices"
hallow thistle
now where would teto, miku, and vtubers + every other (technical e-girl) go? wou...
Sorry if I'm missing something. But why this question?
devout dome
it was when local said if he could he would remove the other e-girl voices
and i more than likely didnt understand what that truely meant cuz isnt an e-girl just a girl on the internet
but then he went further to explain that he meant the "whispery mommy" shit
which tbh is understandable
viral mason
it was when local said if he could he would remove the other e-girl voices
It's like comparing these two
Second image taken from a post here in voice models
First one is cool and very interesting, second one makes my skin crawl
hallow thistle
There's a difference between an "E-girl" voice model that basically sounds like an actual woman and those funny Vocaloid/UTAU voicebanks. Hope this clears up a bit. 
viral mason
While they might both be girls one is definite and real and the other is bait
devout dome
While they might both be girls one is definite and real and the other is bait
hmmm what about people the person who voices the shark girl in ZZZ ( i cant remember the name) would that be in the second or would that still be in the silly area?
hallow thistle
viral mason
hmmm what about people the person who voices the shark girl in ZZZ ( i cant reme...
Just look up girl in the voice model section, you'll see the difference in what I'm talking about
Who is this shark girl
devout dome
hold on lemme find
hallow thistle
Who is this shark girl
More like Ellen Joe from Zenless Zone Zero.
devout dome
More like Ellen Joe from Zenless Zone Zero.
YEAH HER
really? oh you know that does make sense now that i think about it
delicate cradle
yea
devout dome
all the vids of people going in cs lobbies or other games with a "e-girl" voice changer
delicate cradle
we used to have a commissions channel and some people would like center their "shop" around egirl voices
the old server owner also made a egirl voice
in #1175430844685484042 there was a egirl tag
viral mason
Ick
delicate cradle
a now very bad and not properly made pretrain was made to make them better
viral mason
we used to have a commissions channel and some people would like center their "s...
This makes me sick ❤️
delicate cradle
This makes me sick ❤️
you used to be chill with it 🙏
devout dome
mmm ok i was just curious cuz it was reminded to that egirl voices were looked down upon
viral mason
mmm ok i was just curious cuz it was reminded to that egirl voices were looked d...
By most of the server yea
delicate cradle
Proove this vile claim
maybe you didnt say it but you certainly didnt have this much distain for them
viral mason
maybe you didnt say it but you certainly didnt have this much distain for them
The many many weirdos joining and asking over and over to troll has made me dislike them more
They are all the same too, no personality
Just freaks
devout dome
well one more question before I go that isnt a moral one
is it better to use models with a high amount of epochs or low (high= >150 low<= 100) i dont really know where it truely is high or low so im just airballin it
devout dome
what does it mean?
viral mason
Just how many times the model saw the dataset until it sounded good to whoever made it
devout dome
oh thats actually pretty cool
viral mason
A model could sound better at 100 but somehow get worse at 130
And then better again at 150
It's funny
devout dome
ohh like that ai walking experiment where that starts at gen 1 until whatever and starts walking but then could just start flopping outta nowhere
ok well ty again :D
bronze pier
when I click run-applio.bat how long should it take for the window to show up in the browser?
viral mason
Idk I use applio on kaggle
devout dome
I'm eating breakfast
alrighty imma prolly study for my final exam soon until then
viral mason
Ah school, trauma is already flashing through my head like war
I'll get to business then before I start having a mental breakdown
devout dome
Ah school, trauma is already flashing through my head like war
ye im studying mechanical engineering so it truely is hell
devout dome
lol
bronze pier
How do i get it to read the dataset in this path? I am using https://colab.research.google.com/github/iahispano/applio/blob/main/assets/Applio.ipynb and running on public URL
viral mason
How do i get it to read the dataset in this path? I am using https://colab.resea...
<@&1159293204038955078>
make sure to use that ping if you need help quickly
bronze pier
ah thank u
stable basalt
@viral mason is there tutorial for ut
viral mason
no all tutorials on yt are outdated and do not use the new stuff
for vonovox just run start
and vac lite run setup64 then install driver
night condor
heyyy
night condor
yehh'
viral mason
what with?
rugged thicket
Hey, can I ask about co-founders here?
night condor
what with?
Hey guys, do you know any free AI detector and humanizer tools? Not paid one tho,
viral mason
oh sorry no
I don't maybe someone else does tho
@soft karma what in particular are you looking for in rvc?
I saw ur messages in ai chat
soft karma
yeah, just asking tbh. I don't know if anything really changed or if is there another model like it
it's just fun to hear specific characters singing or saying something i want
i've heard the TTS space is getting a lot, but not sure if what rvc does have a better alternative
viral mason
only new improvements really is applio being what is used for training, two new good pretrains both being legacy core 1.5 and 1.6 as well as the pabp which is also decent, as well for realtime Vonovox is the best (Nvidia only)
frail ledge
@viral mason #1175430844685484042 How can I access this channel?
viral mason
<@1023278814752677918> <#1175430844685484042> How can I access this channel?
can you not view it?
frail ledge
can you not view it?
It's not showing up; nothing happens when I click on it.
soft karma
only new improvements really is applio being what is used for training, two new ...
hm, i see. Not much then
still, where can i see all that? want to fiddle with it a little
frail ledge
@viral mason Can you help me? I can't view the channel.
viral mason
<@1023278814752677918> Can you help me? I can't view the channel.
idk what would cause that, did you try closing discord and reopening?
viral mason
hm, i see. Not much then
still, where can i see all that? want to fiddle with it...
how exactly? I'm unsure what you're looking to use
frail ledge
idk what would cause that, did you try closing discord and reopening?
I turned the computer off and on again and it fixed the problem.
frail ledge
idk what would cause that, did you try closing discord and reopening?
Thank you
soft karma
how exactly? I'm unsure what you're looking to use
what rvc does, basically
speech conversion
change voice of speech/singing based in a voice model/reference
viral mason
it can be ran either locally on pc or on browser like kaggle or google colab
soft karma
yeah, i remember using it in the past. Even training some models. Just wanted to ask if anything better released since then
rapid echo
how do i fix Applio port issues i just get "Failed to launch on port 6969" etc
delicate cradle
yeah, i remember using it in the past. Even training some models. Just wanted to...
rvc is still the best speech to speech out there sadly
viral mason
how do i fix Applio port issues i just get "Failed to launch on port 6969" etc
are you using it locally?