How much batch size for smaller datasets of 2-6-10 minutes?
Is it always 8 or I should go lower?
#✨│ai-help
brittle wing
4 or 8?
brittle wing
The only way is by reading our guides
Can you just tell me what batch size do I pick for my dataset length cause I have to train models with small datasets sometimes
simple ore
4 <30m
simple ore
yes
AI HUB Docs
timid olive
This place creates confusion in my mind, that's why.
echo totem
i downloaded rvc on my new pc and using the same voice models as before, but when i try them theres extreme delay and splitting happening which didnt happen on my old pc.
i would show my rvc config if i had image perms.
also yes my correct gpu is being used and i messed with the chunk and extra options.
timid olive
I used to say it was always the lowest, but this is what made my head spin.
brittle wing
This place creates confusion in my mind, that's why.
How do you see that stuff what do you use
timid olive
How do you see that stuff what do you use
applio
brittle wing
This place creates confusion in my mind, that's why.
Stop bugging me please
timid olive
local
timid olive
Yeah, okay, but Colab also provides this information.
I'm doing it from local for the first time right now, I normally used to do it from Colab.
This information was also written in Colab.
@knotty moth Buddy, how can I tell if it's overfitting?
2. The second image confused me. I was always taking the lowest point.
I'm using Applio from local. It cut off for a while, I restarted it, and it's continuing from where it left off.
brittle wing
Whatz goin on
brittle wing
<@681186927151546397> Buddy, how can I tell if it's overfitting?
2. The second...
Did it go up after 90k?
8 something is like 90 but in smaller numbers maybe
timid olive
Did it go up after 90k?
brittle wing
Keep training
brittle wing
Wait WHAT
These were in SVC! They're back
Well to check how real your model sounds!
timid olive
Well listen to them while you're training
Can’t I replace that sample sound with another sound?
coral vault
is there any way to make rvc spew out stereo audio instead of mono audio because im having a hard time finding a way to increase stereo width
gentle rune
help pls if I install cable I can't hear discord anymore and when I delete it my sound comes back I don't know what to do I feel like I've done everything
dusty patio
hello, i'm super newbie here, i keep wathcing everyone said "Google Collab Broken" and using Huggingface instead, so any tutorial using hugging face? sorry for my broken english btw. i really want to be part of this comunity
rain cedar
yoo
proven hill
@low shard the space should be fixed in a bit
viscid moss
Fixed!
viscid moss
<@911742715019001897> the space should be fixed in a bit
Hewwo
u found a workaround? 👀
proven hill
Hewwo
u found a workaround? 👀
someone found a fix for both ilaria rvc and ilaria upscaler, the problem was a git
viscid moss
:0
proven hill
yea it was outdated
vernal hawk
hello guys! does anyone know how to convert e.g. drake vocals into kendrick lamar vocals? i saw it in some posts in the voice-models channel. someone here helped me already setting up the wokada RVC but if i wanted to record some vocals with my voice and then convert them into another voice (basically not RVC, just post-converting), how would i do that?
also i was told to ask this question here again, because you cannot do that in wokada.
proven hill
#📰│dev-updates message @viscid moss @low shard
viscid moss
https://discord.com/channels/1159260121998827560/1159380240271953940/13306276266...
weeeeee
viscid moss
That improvement in time when loading the repository is great!
ty
viscid moss
Both es and english right?
yep
low shard
Alr
low shard
https://discord.com/channels/1159260121998827560/1159380240271953940/13306276266...
Oh wait since when u can do dev updates ?
proven hill
Oh wait since when u can do dev updates ?
since forever
low shard
since forever
Did you ask Vijay for the perm?
proven hill
i always had those
low shard
i always had those
Because technically you don't have any staff role
So wdym since forever
Do all model masters have that perm?
proven hill
no, only me
low shard
no, only me
That's weird since you shouldn't be able as you don't have any staff role?
Is it the kitten role?
proven hill
its added as personal pretty sure
crimson depot
ty
[~] Starting AI Cover Generation Pipeline...
[~] Downloading song...
Traceback (most recent call last):
File "/content/FIX/src/main.py", line 281, in song_cover_pipeline
orig_song_path, vocals_path, instrumentals_path, main_vocals_path, backup_vocals_path, main_vocals_dereverb_path = preprocess_song(song_input, mdx_model_params, song_id, is_webui, input_type, progress)
File "/content/FIX/src/main.py", line 171, in preprocess_song
orig_song_path = yt_download(song_link)
File "/content/FIX/src/main.py", line 76, in yt_download
download_path = ydl.prepare_filename(result, outtmpl='%(title)s.mp3')
File "/content/FIX/.venv/lib/python3.10/site-packages/yt_dlp/YoutubeDL.py", line 1396, in prepare_filename
filename = self._prepare_filename(info_dict, tmpl_type=dir_type, outtmpl=outtmpl)
File "/content/FIX/.venv/lib/python3.10/site-packages/yt_dlp/YoutubeDL.py", line 1367, in _prepare_filename
filename = self.evaluate_outtmpl(outtmpl, info_dict, True)
File "/content/FIX/.venv/lib/python3.10/site-packages/yt_dlp/YoutubeDL.py", line 1358, in evaluate_outtmpl
outtmpl, info_dict = self.prepare_outtmpl(outtmpl, info_dict, *args, **kwargs)
File "/content/FIX/.venv/lib/python3.10/site-packages/yt_dlp/YoutubeDL.py", line 1151, in prepare_outtmpl
info_dict.setdefault('epoch', int(time.time())) # keep epoch consistent once set
AttributeError: 'NoneType' object has no attribute 'setdefault'
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: 'NoneType' object has no attribute 'setdefault'
viscid moss
[~] Starting AI Cover Generation Pipeline...
[~] Downloading song...
Traceback (...
Hmmm
I think it's the video link, it's region restricted or contains sensitive content.
hasty folio
can anyone help me get a good voice changer i tried with yt and stuff but it didnt work
azure marshBOT
can anyone help me get a good voice changer i tried with yt and stuff but it did...
timid olive
It came to 427.
@viscid moss Nearly 500, but the traffic is still going down
hasty folio
!howtoask
crimson depot
Hmmm
I think it's the video link, it's region restricted or contains sensitive c...
It's a song by Bring Me The Horizon
hasty folio
hello i have been trying to get a voice changer but when i found one it was to hard to understand and wasnt working no matter what i tried so if anyone could help me out with which to get and a good tutorial that would be very helpful thank you
viscid moss
<@274566299349155851> Nearly 500, but the traffic is still going down
Keep training and see
timid olive
Keep training and see
I will stop at 500, should I increase the number and run it again?
will stop by itself
I had 2 interruptions, I ran it. continued from checkpoint
viscid moss
I will stop at 500, should I increase the number and run it again?
Ye i think so
viscid moss
It's a song by Bring Me The Horizon
What's the link?
timid olive
Ye i think so
I gave 23 songs to your UVR5-UI-1.7.1 which worked very well for me.
viscid moss
https://youtu.be/Jjl2aomHrLQ?si=yvVd4J4vE0SaCR1n
hmmmm
viscid moss
https://youtu.be/Jjl2aomHrLQ?si=yvVd4J4vE0SaCR1n
Ye it's that video. I tried this one and it works
https://youtu.be/W7itaTpdImE?si=W9HMPVi-RnCeaKWe
crimson depot
Ye it's that video. I tried this one and it works
https://youtu.be/W7itaTpdImE?s...
It's strange. It's the first video that me does that
viscid moss
ye I'm not completely sure what happen
ashen zenith
hiii whats the easiest way to change pre recorded voice recording into any RVC model online on my macbook M1?
brittle wing
should i ask for model making related help on #📑│making-models or here?
crude flame
should i ask for model making related help on <#1159290038115389571> or here?
ask for model making help in #📑│making-models
runic dew
can you guys recommend a female voice model for screaming/crying/shouting ? I'm trying to change my crying voice over from male to female but so far none of them worked because most of the models didn't have any screaming samples while trained.
crude flame
can you guys recommend a female voice model for screaming/crying/shouting ? I'm ...
there are none
rvc limitations
runic dew
there are none
wow .. that's insane
runic dew
rvc limitations
no it's just the dataset
I trained Jack Krauser from RE4 Remake and included his screaming in the dataset , it can detect and transform scream and shouting pretty accurately ..
crude flame
no it's just the dataset
I trained Jack Krauser from RE4 Remake and included his...
yeah but you need to have screaming and such in the dataset and enough of it and if you dont it cant really do any of that
analog obsidian
Its possible to get accurate screaming and shouting, ive seen it in singer models
About crying i dont believe it would work because the current f0 estimations cant correctly estimate the pitch in crying audio
Rvc is kinda mid in cloning expressions, after all, its not a true voice cloning ai
runic dew
yeah but you need to have screaming and such in the dataset and enough of it and...
Yeah that's the real issue
ashen zenith
I have a problem with w-okada
dull steeple
hey, i have an rtx 2070, and an i5 8600k, i should have the right rvc version for my gpu, but it uses only 2% while using 35% cpu
ashen zenith
why does my terminal get stuck on this when i try to open w-okada for first time on my mac M1?
ashen zenith
oh yes thx
stiff goblet
-colab
ashen zenith
this is my version
this is the version i have
no one is replying in #🔍│help-w-okada
wise escarp
did weights go down
uhmmm my "weights" site is showing some weiiird errors
nvm its back up again
dull steeple
i switched to the beta and everything is running pretty nicely rn idk what happened
still not using gpu tho
or well, it's using 0.1% gpu
graceful fiber
do people still use the ULTIMATE voice extractor program or is there a new one? The Guide on the documentation appears to be outdated as the link it gives for voice extraction is 404 NOT FOUND
simple ore
link to a specific model
crude acorn
fiery magnet
how i download a model
steel vine
how i download a model
- voice-models in discord channel
- search model
- Click on the one you want to choose
- there is a link, go to it and download
5 create any folder (where you open start.) or in another place (as is more convenient)
6 from the zip file, transfer both files there
7 click edit on the right under all models in the application
8 scroll down and add your model, imput add imput and another one in the first slot
fiery magnet
1. voice-models in discord channel
2. search model
3. Click on the one you want ...
so i take the shi from the download and ad it to a zip file?
steel vine
so i take the shi from the download and ad it to a zip file?
after downloading the file, in the application, there is editing, under the models, click on it, scroll down and you need to add both files that you downloaded to 2 windows
fiery magnet
1. voice-models in discord channel
2. search model
3. Click on the one you want ...
which rvc should i be using anyways
im using illaira
steel vine
im using illaira
below are there free slots for models or are you talking about something else? below where there is nothing, you can use any slot
hallow thistle
fiery magnet
below are there free slots for models or are you talking about something else? b...
im where it says model loader
and the url wont work
wait
i think i got it
hallow thistle
steel vine
after you follow the link and download the file itself, there will be 2 files inside, go to edit and scroll down, click upload and add both
index - .index
model - .pth
and click upload
hallow thistle
after you follow the link and download the file itself, there will be 2 files in...
Thank you for your answer. But I already know all of that.
steel vine
<:true:1159359915610083348>
Well, you sent a photo, didn't you, for the person above?
I just added to the message, telling what to do next((
stiff goblet
-train
low shard
@nocturne mural check this out
simple ore
um
you did not clone the actual application? 🙂
that's what it looks like.. cloned the app.py, but not the rest
hallow thistle
Just only the app.py file? Where is the rest of repository? 
tame mica
you did not clone the actual application? 🙂
no idea
i didnt change anything except the ngrok token
simple ore
you ran the repo cloning cell, right?
tame mica
yep
hallow thistle
You're looking for a "voice changer" program called W-Okada, go to #🔍│help-w-okada for more information.
raw summit
You're looking for a "voice changer" program called W-Okada, go to <#11592901616...
Oops sorry Ty
gritty forum
@gusty kestrel
median bane
Что делать если ужасный звук с помехами, что можно попытатся сделать?
steel vine
Что делать если ужасный звук с помехами, что можно попытатся сделать?
It depends on the video card, if you have a weak one, then you will have an increasingly strong delay, if you make the delay greater, then the voice will be heard better or the problem is with your microphone
steel vine
i think thats good
On the right in the menu, there are values, extra and chunk, you need to poke them and check the voice, the lower the delay, the better the voice, on the right above the photo of the model, there are 3 types of ms
steel vine
i think thats good
also if you use a female voice, it is advisable to do pitch 11-15
ebon tulip
I'm upgrading to a new build with an RTX 5080 and an i9-14900KF. Will this new setup result in a significant improvement in sound quality over my old 2060S rig?, or does it not really affect it?
knotty moth
I'm upgrading to a new build with an RTX 5080 and an i9-14900KF. Will this new s...
performance ≠ quality, do you mean for training? if not, your 2060 is enough
regarding the upgrade plan, I'd recommend 5090 or 5070 Ti/4070 Ti super, and ryzen 9800X3D depending on ur budget
oblique viper
where are the programs used to train and try models?
ebon tulip
performance ≠ quality, do you mean for training? if not, your 2060 is enough
I have already sold that 2060S rig and purchased the processor and other components which mean switching to 9800X3D is not possible, Now, I’m just waiting for the graphics card, which should arrive in two weeks.
I’d definitely be open to using it with TRAINING, even though I haven’t tried it before. However, when I used Odaka months ago on my 2060s, the sound quality experience was pretty disappointing. I’m curious if this upgrade will make a noticeable difference. For example, could it allow for higher quality settings, which might increase latency, while the upgraded hardware reduces that latency, which will lead to a better overall performance? Also, is there a specific reason why the 5070TI or 4070TI is recommended over the 5080?
low shard
I have already sold that 2060S rig and purchased the processor and other compone...
The quality heavily depends on the model, your gpu will make so you can handle better the latency even while running games
also be sure to get the wokada deiteris fork which is more optimized, we can talk about that in #🔍│help-w-okada
so no it won't make quality better
oblique viper
can someone tell me where I can download the programs for training AI models or atleast for using the models?
low shard
can someone tell me where I can download the programs for training AI models or ...
what's ur pc gpu
and do u mean use them on pre-recorded audios or realtime for calls
ebon tulip
The quality heavily depends on the model, your gpu will make so you can handle b...
When I check out a model and hear how good the example sounds, then I give it a shot myself, it’s just not close. The sound doesn’t match the example, and that’s what I mean by “quality.” I’m not if it will be better than the original, just that it should at least sound like the example.
knotty moth
I have already sold that 2060S rig and purchased the processor and other compone...
- y'know, intel highend 13-14th gen cpus have bad reputation because of the degradation failure incident (before finally the microcode update), not only that but also power hungry?
- get the fork wokada, the 2.7s extra setting is the optimal quality and 3060 can still use that
low shard
- y'know, intel highend 13-14th gen cpus have bad reputation because of the degr...
btw the 1st point got fixed with microcode, I have one myself
however it's bad that it had that issue, rip old owners
ebon tulip
- y'know, intel highend 13-14th gen cpus have bad reputation because of the degr...
Indeed, the microcode update resolved all the issues, and the power consumption can also be improved with some adjustments in the bios. This is because Intel doesn’t provide the best settings by default, unlike Ryzen
oblique viper
what's ur pc gpu
6700 XT
simple ore
read carefully and follow to the letter
oblique viper
and do u mean use them on pre-recorded audios or realtime for calls
there used to be a program where you could import a model, record audio and it'd put the ai voice on your recording
I haven't used this for a while and it looks like things have changed
oblique viper
https://docs.applio.org/applio/getting-started/installation#amd-gpu-support-wind...
thank you
ebon tulip
<@842582771121651754> I gave you the answer, yet you still went for 14900k
Oh, you mean on the other discord server? I actually bought the processor a few weeks ago before that
simple ore
sad
knotty moth
I have already sold that 2060S rig and purchased the processor and other compone...
tbh it depends on games you want to run, but 5080 still has 16 GB vram like the last gen 4080/super
ebon tulip
tbh it depends on games you want to run, but 5080 still has 16 GB vram like the ...
It's GDDR7
knotty moth
It's GDDR7
and 5070 Ti is also 16 GB, just like 4070 Ti super
ebon tulip
and 5070 Ti is also 16 GB, just like 4070 Ti super
Yeah, all three of these cards are actually 16GB. So, why are the 4070TI Super/5070TI recommended over the 5080?
knotty moth
Yeah, all three of these cards are actually 16GB. So, why are the 4070TI Super/5...
value proposition, unless you can stretch your budget for 4090/5090 that have much more vram
low shard
When I check out a model and hear how good the example sounds, then I give it a ...
did u try playing with the pitch
knotty moth
When I check out a model and hear how good the example sounds, then I give it a ...
for better "quality", just get better model or train the one by yourself
analog obsidian
When I check out a model and hear how good the example sounds, then I give it a ...
This depends in the model ability to generalize, models that generalize bad aren’t able to do male to female conversion and viceversa
fluid cliff
how do i make an rvc of my own voice ?
ebon tulip
value proposition, unless you can stretch your budget for 4090/5090 that have mu...
I think all these GPUs, (RTX 4080, 4070 Ti Super, and others), are priced similarly. Production of the 4080 and 4070 Ti Super has completely stopped, which has driven their prices up. From what I’ve seen on Amazon, they aren’t really cheaper than the RTX 5080. Also, the 16GB VRAM in the 4080 Super isn’t the same as in the 5080, since GDDR7 significantly reduces VRAM consumption, according to what I have read in their official website, (Micron/Samsung)
low shard
how do i make an rvc of my own voice ?
what's ur pc gpu
fluid cliff
what's ur pc gpu
rtx3060
heavy arrow
yo which one of the collabs automatically separated everything for you if you provided a song and made it hella easy
low shard
yo which one of the collabs automatically separated everything for you if you pr...
-colab
low shard
AICoverGen
heavy arrow
bet 🙏 thank you og its been over a year since i messed with rvc
knotty moth
I think all these GPUs, (RTX 4080, 4070 Ti Super, and others), are priced simila...
well have a good luck to find the best price of 5080 then
low shard
bet 🙏 thank you og its been over a year since i messed with rvc
yw and lmk
knotty moth
there's a rumour that 5090 could be scalped much higher than the $2000 msrp (including the import tariff and shit)
ebon tulip
I think all these GPUs, (RTX 4080, 4070 Ti Super, and others), are priced simila...
Also, since I'm using RTX HDR which reduces performance & DLSS transfomer which reduces performance by 5% & DLDSR which reduces performance as well, I really need an RTX 5080 over RTX 5070TI
fluid cliff
As you got a good PC, you can use RVC locally, you can choose between:
- [Applio...
i can make an ai rvc of my OWN real voice with this ?
i mean my actual voice
low shard
i can make an ai rvc of my OWN real voice with this ?
yes, you just need to create a dataset of your voice
ebon tulip
did u try playing with the pitch
Yes, I did, for several days
ebon tulip
This depends in the model ability to generalize, models that generalize bad aren...
Does this mean it's better to train my own model rather than downloading one or I am getting it wrong?
knotty moth
Also, since I'm using RTX HDR which reduces performance & DLSS transfomer which ...
again it depends on the game and the resolution & graphics settings, so you should wait for real benchmarks & reviews before deciding
tropic garden
are there any good videos on applio? I tried to infer but there is no export.
low shard
are there any good videos on applio? I tried to infer but there is no export.
Nope all videos are outdated
analog obsidian
Does this mean it's better to train my own model rather than downloading one or ...
It means the model you’re trying to use is unable to create audio correctly using your voice as input due to being badly trained
tropic garden
don't know if my model is loading
low shard
Could you share ur pc gpu, which tutorial link are u following, and a screenshot?
tropic garden
using google collab m2 chip mac
oblique viper
https://docs.applio.org/applio/getting-started/installation#amd-gpu-support-wind...
is applio also the program where you put a recording and get an AI version of it?
low shard
using google collab m2 chip mac
oh yeah since mac isn't that good you gotta use google colab which is a cloud (remote good pc) optioN
could u share what tutorial link are u following and a screenshot of ur issue
simple ore
I think all these GPUs, (RTX 4080, 4070 Ti Super, and others), are priced simila...
"GDDR7 significantly reduces VRAM consumption" makes zero sense
simple ore
is applio also the program where you put a recording and get an AI version of it...
it is a voice conversion program, you put in audio with Joe Biden's congressional speech, apply a voice model of a Spongebob, the result is Spongebob spewing political nonsense\
oblique viper
perfect that's what I'm looking for
oblique viper
it is a voice conversion program, you put in audio with Joe Biden's congressiona...
is it also used for training models or is that another program?
low shard
is it also used for training models or is that another program?
yup it is
simple ore
same program
oblique viper
awesome
tropic garden
Thanks. I can't find the upload button for the screenshot
storm mica
I did the collab thing and trained a test model and the outputs were a .index file and a .o file. I used the .o successfully once but now it isn't working. The prog (ilaria for example) want a .pth file. I also uploaded it to HF as a model and pointing to the model card (main model URL i believe) also isn't working. Tips?
tropic garden
I am back to Weights. I'll work on that Appliio on my PC. ;).
knotty moth
I did the collab thing and trained a test model and the outputs were a .index fi...
isnt .o a compiled c/cpp object file? and which colab are u using? afaik none of colabs we recommend here have that strange issue
low shard
Thanks. I can't find the upload button for the screenshot
!give-media-perms 30m @tropic garden
you should be able
low shard
I am back to Weights. I'll work on that Appliio on my PC. ;).
I mean your choice what to use
storm mica
isnt .o a compiled c/cpp object file? and which colab are u using? afaik none of...
Looking in my collab RVC disconnected folder I see that and some .pth files. Sorry for the premature question. i need to dig around a bit more i think
humble river
FOR RVC NO ERRORS: USE THE ONE IN RELEASES AND ONLY CLICK GO-WEB.BAT
tropic garden
here you go!
simple ore
here you go!
you did not select the voice model
tropic garden
I don't get a list in the menu under model. I'll try again and refresh and see.
ok it worked now, thanks.
heavy birch
HOW DO I MAKE THE VOICE MODELS IN WEIGHT LOUDER
low shard
@ashen zenith
For Inference (use models) Mac, You can:
- Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides, probably won't be able to train, make models):
- Cloud (remote good pc, easier and faster than ur PC but it's limited):
- Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
- Ilaria RVC Zero: fastest and simplest that you can get for free
- Applio Colab: max 4 hours, not granted, of GPU
sonic fiber
does any of you guys use Vidu ? i have ran out of my free credits for january and i can't wait 10 days to finish my ai movie, can someone make me an account ?
if someone doesn't use vidu at all for videos, they should probably not mind me using all of their monthly free attempts , so just let me knowww
it's called vidu ai
in case you wanna find the website
hallow thistle
Why are you asking this in #✨│ai-help ? This channel here is all about RVC the audio changer and certain RVC programs.
For AI video generator, you can go to #🔍│help-ai-art .
grave ferry
Having trouble with crashes, I am using Applio. Running certain models shuts my whole pc down, leaving me to flip the power supply off and then back on to restore functionality to my power button. I have a Ryzen 5700x3D, 3060ti and 750W PSU.
Can't find the details of the one I have working but its the Playboi Carti (I AM MUSIC Era Deep Voice)[RVCV2 RMVPE] 290 Epochs that I found in the voice models section in here.
Details of the model that crash my PC (Yeat Lyfe Era [RVMPE / 500 Epochs ] RVC V2
Sample Rate: 42k
Epochs: 500
Pitch Extraction: RVMPE
Batch Size: 8
I'm wondering if it's just the Epoch amount? I'm not super familiar with this so i'm not sure if that's a factor.
sonic fiber
Why are you asking this in <#1159290139609137264> ? This channel here is all abo...
its cause im slightly dumb, ill head over there
knotty moth
Having trouble with crashes, I am using Applio. Running certain models shuts my ...
not a BSOD crash right? it seems more likely the faulty PSU than the possible transient spike, as a pc with RTX 4070 and 750 W PSU never had such issue
violet swan
Is g/total 24 good?
hasty sleet
Hi everyone,
I’m not sure why I encountered the following error today while running the project "AICoverGen_colab.ipynb":
raise ValueError(f'mutable default {type(f.default)} for field '
ValueError: mutable default <class 'fairseq.dataclass.configs.CommonConfig'> for field common is not allowed: use default_factory
Timer: 00:11:43
I tried searching for the cause of this error and was wondering if it might be due to Colab using Python 3.11. Any insights?
low shard
Hi everyone,
I’m not sure why I encountered the following error today while ru...
send the link of the colab
grave ferry
not a BSOD crash right? it seems more likely the faulty PSU than the possible tr...
Nah no BSOD, just straight cut power, I was assuming it was PSU related just wanted to make sure I wasn't missing something, thank you!
odd flower
how, can you please share?
hasty sleet
send the link of the colab
hasty sleet
send the link of the colab
I used this Colab, and it was running normally on Christmas Day.
low shard
https://colab.research.google.com/github/SociallyIneptWeeb/AICoverGen/blob/main/...
don’t use that one
-colab
brittle wing
I am extremely new to okada and i have one issue . My mic is not responsive in the app. like if i try sound recorder in my pc, the sound is well picked up but in okada rvc, no input
low shard
I am extremely new to okada and i have one issue . My mic is not responsive in t...
Use #🔍│help-w-okada
idle crypt
Hi! how are you guys!
I trained a model with 50e with AI Hispano NoUI in google colab but I can't seem to resume training I ahve this error... I stopped the training process because it ended and then headed to the "resume training" part.
Any idea of what is the issue?
Error:
"FileExistsError Traceback (most recent call last)
<ipython-input-12-6951881e8132> in <cell line: 0>()
18 if os.path.exists(time):
19 shutil.copy(time, time__)
---> 20 shutil.copytree(source_path, destination_path)
21 print("Model backup loaded successfully.")
2 frames
/usr/lib/python3.11/os.py in makedirs(name, mode, exist_ok)
FileExistsError: [Errno 17] File exists: '/content/Applio/logs/Sam_Eng_Mod_GD_V1'"
agile storm
guys can someone tell me a nice local inference gui please? (not applio pls)
idle crypt
How many epochs are good enough for a model to work with 24 minutes of talking dataset?
fast scarab
Hi,
Why don’t I have the index file when I create my voice functionality?
vocal flower
hi, someone can do model iceicell ?? It’s not working at all, I’ve been sitting for 4 hours
ember wadi
/content/RVC
Traceback (most recent call last):
File "/content/RVC/demo.py", line 1, in <module>
from original import *
File "/content/RVC/original.py", line 3, in <module>
from dotenv import load_dotenv
ModuleNotFoundError: No module named 'dotenv'
Anyone has a RVC -colab version that works??
vague sage
is this right? it says 0%
low shard
Anyone has a RVC -colab version that works??
-colab
vocal flower
is this right? it says 0%
why i dont have this downloading ---------------------------------------------------------------
write filelist done
use gpus: 0
runtime\python.exe train_nsf_sim_cache_sid_load_pretrain.py -e ICEICELL -sr 48k -f0 1 -bs 1 -g 0 -te 150 -se 50 -pg pretrained_v2/f0G48k.pth -pd pretrained_v2/f0D48k.pth -l 1 -c 1 -sw 1 -v v2 ----------------------------------------------------------------------
only this
last line
vocal flower
help pls
cyan torrent
@tired kraken
ancient crow
estou passando dificuldade 💔
knotty moth
<@678457649431314479>
don't ping random person, as he only supports SD/flux related topics in #🔍│help-ai-art , not RVC
also that means you have used different architecture and/or sample rate from what you used in preprocessing & feature extraction
cyan torrent
don't ping random person, as he only supports SD/flux related topics in <#118567...
when i use the normal pretrained it doesnt give me any errors
but when i use DMV Pretrained it gives me that error
knotty moth
when i use the normal pretrained it doesnt give me any errors
but when i use DM...
no it is default pretrain in 48k
cyan torrent
no it is default pretrain in 48k
wym
when i use this
https://discordapp.com/channels/1159260121998827560/1329927130494144622/1329927130494144622
pretrained it gives me that error
but when i use the normal pretrained/default one it doenst give me any errors
knotty moth
when i use this
https://discordapp.com/channels/1159260121998827560/132992713049...
it is 32k like the previous v1 https://discord.com/channels/1159260121998827560/1255206706598772796
cyan torrent
it is 32k like the previous v1 https://discord.com/channels/1159260121998827560/...
alright
thx
t
hallow thistle
-rvc
hallow thistle
do you remember sokar?
No.
narrow dock
No.
that makes me sad
none rememberd me when i made the most famous voice models here
its probably still used
20000 CREATIONS WHAT
bruh im Famous
it was worth it deleting my account
An unexpected error occurred: No connection adapters were found for ':https://huggingface.co/yaniroo/YanAI/resolve/main/MikuAI.zip'
how do i fix it?
when i try to download a model
civic kettle
How do I use the Ai?
hallow thistle
How do I use the Ai?
!howtoask
silent stratus
How do I use the Ai?
install applio or use weights.gg
silent stratus
``An unexpected error occurred: No connection adapters were found for ':https://...
try a diff client or u can prob use weights.gg
rancid sluice
!help
dull ironBOT
tawdry gullBOT
!help
LunaBot 🌙 is the perfect music bot! Feature rich with high quality music! And Custom Playlist
You can start listening music by just joinning a voice channel and typing: /play [song name or link] (Remove brackets).
We support only Spotify, soundcloud, bandcamp and more!
To view more help on a specific command or category, run
/help <command> or /help <category>
Important Links:
Support
Premium
Invite
Command Categories:
🎶: Music
💰: Premium
⚙️: Utility
📕: Admin
Select A Page From Dropdown Menu Below
LunaBot
rancid sluice
-rvc
azure marshBOT
latent cypress
should i enable the warmup phase when i resuming my training?
simple ore
warmup is only for initial training start in codename's fork
fiery juniper
is it just me or weights.gg doesnt give index files anymore
IM CURIOUS... I WANT TO GET INTO MODEL MAKING
HOW DO YOU GUYS MAKE MODELS 😭
hallow thistle
To get help about training a voice model, you can ask anyone in #📑│making-models. They know how to train a voice model either locally or on Colab/Kaggle.
A Weights-trained voice model with no index file can happen when the dataset audio is either short or something else. RVC system in Weights is automatic, anything will set up by itself, so you won't be able to force it to train an index file other than the dataset audio you uploaded to there.
@ripe raven What's up? I see you're trying to get into my direct message.
ripe raven
My voice doesn't change in the program, it's not like anime.
ripe raven
<@1138367543975882782> What's up? I see you're trying to get into my direct mess...
Sorry I just wanted to talk to you about my problem and how I can make my voice like anime.
Can you help me?
azure marshBOT
Can you help me?
hallow thistle
Sorry I just wanted to talk to you about my problem and how I can make my voice ...
Go to #🔍│help-w-okada.
ember wadi
@low shard Thanks man, the RVC disconnect/ RVC Mainline neither works, but i found the problem, it was colab upgraded python 11 few days ago. there were dependency issues
knotty moth
@ember wadi no you aren't the thread poster and ur topic is different from it at all, since it can't be moved like in most forum sites, I've deleted yours there and repost it here.
Oh! Seems here is the right place to ask question!
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/wiki/FAQ-(Frequently-Asked-Questions)#q9how-many-total_epoch-are-optimal
Q9:How many total_epoch are optimal?
If the training dataset's audio quality is poor and the noise floor is high, 20-30 epochs are sufficient. Setting it too high won't improve the audio quality of your low-quality training set.If the training set audio quality is high, the noise floor is low, and there is sufficient duration, you can increase it. 200 is acceptable (since training is fast, and if you're able to prepare a high-quality training set, your GPU likely can handle a longer training duration without issue).
Can anyone tell me why RVC takes only 20-30 but Applio needs 200-300? if 20-30 epoches can do the work it might avoid overtrainning, right??
fathom reef
Is Ilaria bugged out or something?
Because I tried converting my voice into Yoshi's voice (don't ask why,) and it decided to use a completely random voice instead. Whoever's voice it is, I have never heard of it.
I'll try and get some footage in.
What it sounds like on Ilaria:
VS what it should actually sound like:
low shard
is it just me or weights.gg doesnt give index files anymore
Try reporting it in their server
fathom reef
Is Ilaria just broken rn?
knotty moth
<@681186927151546397>ok thankd
answering to ur question, you need to analyze the tensorboard graphs as they may vary on every audio quality, pretrain used, and other unknown factors, and also test some interesting model checkpoints
low shard
<@911742715019001897> Thanks man, the RVC disconnect/ RVC Mainline neither works...
Both should be fixed now
fathom reef
I tried it with W-Okada, and it sounds exactly like it should.
But with other RVC voice cloning software like Ilaria, it sounds extremely off.
fathom reef
What it sounds like on Ilaria:
Play the footage to see what I mean.
fathom reef
Still waiting on some help with this.
fathom reef
Yeah, I know! It's really weird!
proven hill
Yeah, I know! It's really weird!
let me restart the space, if you can try again in 10 minutes
fathom reef
Cool! I'll check back.
proven hill
make it 5 actually
also, try to use index to max, could be that maybe(?) im kinda clueless
fathom reef
Nothing.
worn prawn
When the training on the applio ends, where does all the data go and how do I download it?
fathom reef
let me restart the space, if you can try again in 10 minutes
It's fixed. Thanks!
proven hill
It's fixed. Thanks!
happy to hear that!
simple ore
When the training on the applio ends, where does all the data go and how do I do...
it says right in the log
worn prawn
and of course /logs folder
I am doing the training on the website, but does it automatically create a folder called logs on my computer?
simple ore
if you're using colab, then it shows a log for each cell
worn prawn
yes colab applio
pearl cove
Does anyone know why my shit is getting cut off and shit after the conversion cuz It didn’t happen before but now it’s just like cutting out getting quieter and the voice kinda off it worked perfectly before
I use Mangio rvc fork
odd shale
I use Mangio rvc fork
Are you using Mangio Fork's built in realtime?
I would suggest you to try deiteris' fork instead.
Also, what's your GPU?
hallow thistle
Mangio as an RVC fork program is considered outdated.
pearl cove
Are you using Mangio Fork's built in realtime?
Web gui thing
pearl cove
Also, what's your GPU?
3060 ti
odd shale
Mangio as an RVC fork program is considered outdated.
Yep, that's right.
pearl cove
Idk what happened tho worked perfectly b4
odd shale
Idk what happened tho worked perfectly b4
But what did you try to do? inference with prerecorded audios or realtime talking?
Also yep, Mangio is pretty outdated
pearl cove
But what did you try to do? inference with prerecorded audios or realtime talkin...
Recorder audio
odd shale
Recorder audio
So, inference?
odd shale
Put my audio in select model boom convert
Hmm, i see. It's inference then.
Welp, i think you should install Applio or RVC Mainline instead
Because Mangio is really outdated since 1 year and half
pearl cove
Where the links at
pearl cove
Preciate u
primal bridge
Hi can u guys help me in setting up real time voice changer for valorant?
hallow thistle
Hi can u guys help me in setting up real time voice changer for valorant?
For W-Okada, go to #🔍│help-w-okada.
primal bridge
For W-Okada, go to <#1159290161683767298>.
What is W okada btw?
I am new here so idk much
hallow thistle
What is W okada btw?
W-Okada is the realtime voice changer.
worn prawn
I just completed the training, and I can't find where the logs went
fathom reef
Is it made so that it requires the EPOCHS to be the exact same as well??
They're similar enough!
I think it actually does this with every single model you use.
low shard
Is it made so that it requires the *EPOCHS* to be the exact same as well??
Are you using Ilaria RVC Zero?
analog obsidian
Is it made so that it requires the *EPOCHS* to be the exact same as well??
Both models have to match the sample rate
So for example if one model is 40k and the other is 32k, merge will not be possible
Edit: i didnt saw the screenshots lmao 😭
low shard
iirc merging models is also broken on ilaria rvc zero
fathom reef
Both models have to match the sample rate
So for example if one model is 40k and...
That's exactly the issue: They are the exact same sample rate.
fathom reef
Are you using Ilaria RVC Zero?
Noted. What other non-local alternatives are there for model fusion?
low shard
Noted. What other non-local alternatives are there for model fusion?
Maybe you could try Applio UI colab
or mainline, or applio no ui colab, however not sure if the last 2 got this feature
fathom reef
Yeah, WebUIs are against Google's Terms of Service, sadly.
low shard
@proven hillbtw the model fusion thing could just be removed if it's still broken in ilaria rvc zero
low shard
Yeah, WebUIs are against Google's Terms of Service, sadly.
we use encrypted code
they do not get detected
so u won't get banned
people do this since like 2 years almost
fathom reef
Where is the link to this one, then?
analog obsidian
low shard
Where is the link to this one, then?
the bot isn't working rn but u can check #📰│dev-updates message
indigo olive
https://colab.research.google.com/github/iahispano/applio/blob/master/assets/App...
how do i use it?
its a tutorial on youtube?
simple ore
They're similar enough!
as long as the parameters are the same, you can merge them
low shard
its a tutorial on youtube?
Nope all YouTube tutorials are old
Could you tell me what u want to do and what's ur PC GPU
indigo olive
where i put the pth voice @low shard
low shard
where i put the pth voice <@911742715019001897>
answer my questions so I can help you
azure patio
guys i have my own models and i wanna make covers with them but which would be the best colab notebook to use based on personal experience
glad zealot
guys i have my own models and i wanna make covers with them but which would be t...
Mainline for actually changing the voice and uvr colab for splitting the vocal and background
hallow thistle
hallow thistle
where i put the pth voice <@911742715019001897>
On Applio, you go to this Download tab.
And there's a place where you drag-and-drop pth and index files. That's where you upload the voice model you downloaded to Applio.
Voice model glitching on W-Okada can occur when you set chunk and extra settings way too high for your GPU or W-Okada is just lagging hard in your system.
For more information about W-Okada, go to #🔍│help-w-okada.
foggy basin
what does everyone use to separate vocals and instrumental for music? im using a website right now thats the best i can find but it has a limit of 1 per day and i cant be bothered to pay $5 a month to make shit posts
hallow thistle
what does everyone use to separate vocals and instrumental for music? im using a...
The UVR5 is used to seperate audio into stems.
knotty moth
what does everyone use to separate vocals and instrumental for music? im using a...
there is a complete guide: https://docs.google.com/document/d/17fjNvJzj8ZGSer7c7OFe_CNfUKbAxEh_OBv94ZdRG5c/edit?tab=t.0
you can ctrl+F what you want to look for
foggy basin
there is a complete guide: https://docs.google.com/document/d/17fjNvJzj8ZGSer7c7...
i think im fine i dont have the time or comprehension to dig through a 456 page manual to make more than 1 song a day ill be fine with the website wow thanks for trying though i just use rvc gui 2 to make stupid music im not a master here sorry
lavish lintelBOT
foggy basin
https://cdn.discordapp.com/emojis/1015980727227138128.webp?size=48
this is why i dont interact with the ai community on discord tf is that
knotty moth
i think im fine i dont have the time or comprehension to dig through a 456 page ...
why still using that old tiger18n's repo?
foggy basin
because it works...
foggy basin
im not a master here dont expect me to be
hallow thistle
The bot is not working. 
foggy basin
i just make funny things with a gui that i found
knotty moth
hallow thistle
RVC-GUI is too old now.
foggy basin
idk man dont bash me i use it and it sounds just fine...
im not going to install 30 things to make a video game character sing a funny song
foggy basin
im not blaming myself dude
all 500,000 come running
to help some guy make spongebob say the n word
alot of servers have idiots that just come in and do things like that
knotty moth
because it works...
"if it ain't broken don't fix it" till you found that the latest RVC/applio has better quality than the old rvc gui you're using
foggy basin
why the hell are you so hostile
i dont really notice any difference in sound quality me vs the average person who rewrote their entire system and bought 30 things to make slightly better ai covers
knotty moth
because it works...
and inference using rmvpe is better, if you can't find it in yours, better ditch it
foggy basin
ah ok?
hallow thistle
why the hell are you so hostile
Are you telling me I'm so harsh about it? What???
foggy basin
not you
knotty moth
to help some guy make spongebob say the n word
sorry but anything with n words are strictly prohibited in this server
foggy basin
do any of you have common sense here oh my god im done with this conversation it feels like im talking to an old folks home lecturing me for not driving an electric car because the news said it was good
foggy basin
sorry but anything with n words are strictly prohibited in this server
are you legit stupid
i didnt say i did it
i said that some guy who would would @ everyone and just ask how to do stupid shit like that
foggy basin
how much r34 have you consumed to rot your brains this bad it feels like im talking to a wall dude
knotty moth
are you legit stupid
well nothing forbids you to do anything for yourself, though it's your responsibility
hallow thistle
What? Are you alright? No, I ain't argue with you either, and you acted like you're so confident about yourself enough to get me wrong.
I do what I think it's normal to do. Anything you say is like you know things better than me.
foggy basin
What? Are you alright? No, I ain't argue with you either, and you acted like you...
what the hell are you talking about man
I NEVER SAID YOU WERE BEING RUDE OR ARGUING HERE just to clear that
im not talking aobut you
I was talking to the other dude
im not acting confident i dont know anything here i came here acting like your peasant since your more experienced at this type of thing asking for advice and i kindly said that no im not reading all of that to change a small things and you act like im some kind of neanderthal
because i use a slightly old gui here
hallow thistle
You straight up asking the other if he's actually stupid. The "actually stupid", he's just like that as a normal person.
foggy basin
neither of you are normal
foggy basin
You straight up asking the other if he's actually stupid. The "actually stupid",...
stop dick riding him
im out of here so you dont get offended and ban me or something i like the server
hallow thistle
neither of you are normal
Then I don't think you're normal either.
foggy basin
ok cool bud👍
you talk like your 8 man are you a non native english speaker? or something
hallow thistle
And you're gonna blame the AI community for talking shit to you. Oh, no, you just made it up yourself like you're right about everything lmao.
knotty moth
you make me actually sick
hallow thistle
If you still feel like this conversation doesn't feel normal for you, simply just get out, don't try wasting your time for this. Or you can go ask some people who know better than me.
foggy basin
And you're gonna blame the AI community for talking shit to you. Oh, no, you jus...
what the hell are you talking about man
i am not blaming the ai community
what do you have mentaly different here you dont talk like a regular person
ooh your so big and strong for warning me
ok discord mod
i am going to go live a normal life now while you rot away watching anime and listening to abc songs or something because you act 4 years old
hallow thistle
Asking for a tool that can seperate audio into stems
Saying I don't know about AI
Completely loses it, and then straight up calling someone stipud
oak plank
Let's end this topic here, and let's be nice to each other
foggy basin
i have nothing against you weights
i was never mad at you
i never said you dont know about ai
i never lost it and i said HE was acting supid
for calling me a moron because im not up to date with my rvc
latent cypress
is it mean we can't do inference with mrf hifigan model in older fork? maybe i just misunderstood, because i can't understand english very well
modern oxide
for calling me a moron because im not up to date with my rvc
bro, no one ver called you a moron
foggy basin
weights just misunderstood this whole thing
yeah they did
just stfu your not in this dude mute me idc stop making it worse
modern oxide
they just suggested a better thing to use, and you got mad for some reason, chill, no one here is calling you an idiot
hallow thistle
knotty moth
is it mean we can't do inference with mrf hifigan model in older fork? maybe i j...
the mainline/older rvc only supports models trained on the default hifigan
modern oxide
what does everyone use to separate vocals and instrumental for music? im using a...
you can download uvr5 if you want to separate voices locally, or use mvsep.com which hosts the same models, but is limited by 10 minutes long and 100 mb size audio (but is free at least). to use rvc models you can download applio, it's not hard to do, no 20 things to download, it does it better than rvc gui, cus uses new stuff, but works with the same principal, you input audio, choose a model and it gives you audio
latent cypress
the mainline/older rvc only supports models trained on the default hifigan
any clouds notebook that support inference with different vocoders?
knotty moth
any clouds notebook that support inference with different vocoders?
use the applio inference itself
foggy basin
they just suggested a better thing to use, and you got mad for some reason, chil...
they were very clearly saying what i used was inferior and outdated ill cite a few messages
foggy basin
RVC-GUI is too old now.
.
foggy basin
why still using that old tiger18n's repo?
.
your actually helpful here man thank you john
sorry for what i said to you
i have nothing against you
modern oxide
they were very clearly saying what i used was inferior and outdated ill cite a f...
yeah, but it's not really an insult, it's just how it is, it's like comparing 2080 ti and 5090, applio made many cool advancements when it comes to inference speed quality, etc
foggy basin
i dont really care about speed or quality i just make funny things for my friends
but thanks
ill try it
modern oxide
sorry for what i said to you
yeah no problem bro, no one here tried to insult you or anything, they were just recommending better stuff
modern oxide
nah, trust me, they just wanted to help
hallow thistle
.
That is not even an insult. I don't know what made you think like that, but this is how I usually tell people if they use an older program that's way outdated.
foggy basin
That is not even an insult. I don't know what made you think like that, but this...
not that man
these emojis arent helping me man...
i have work tommorow i dont need to deal with this
i would prefer to just talk to john here hes very helpful and kind
i think i have what i need now
thanks
modern oxide
np, but they were also just trying to help
at least rtx 5090 fe was, weights by namari just goofing around
hallow thistle
I can be harsh sometimes, depends on how I talk. But like not everything should be taken way too serious, you know.
foggy basin
i might just not be understanding tone well because english isnt my native
im a native german
modern oxide
i understand, i also had troubles understanding people when i was just starting to learn english (i'm also not a native speaker)
knotty moth
at least rtx 5090 fe was, weights by namari just goofing around
uh u dont think my server nickname is funny? 
modern oxide
uh u dont think my server nickname is funny? <:Baffled:1159397286833557566>
it is funny
modern oxide
uh u dont think my server nickname is funny? <:Baffled:1159397286833557566>
i wanted to make a joke "wow we have 5090 in ai hub before release" in staff chat but just laughed it off in my head lmao
modern oxide
btw, https://youtu.be/RDr1pr_c6ts
this is a marvel of engineering
hallow thistle
btw, https://youtu.be/RDr1pr_c6ts
This GeForce RTX 5090 size looks thinner than RTX 4090.
modern oxide
This GeForce RTX 5090 size looks thinner than RTX 4090. <:Baffled:11593972868335...
it is thinner, it's a 2 slot card, while 4090 is a 3 slot card
hallow thistle
🔥
foggy basin
where do i put my rvc voice .zip files with applio
you can use regular weights models .zip right?
basicially how do i pick what model i want to use with applio
modern oxide
you can use regular weights models .zip right?
yes
modern oxide
basicially how do i pick what model i want to use with applio
in applio folder in logs folder you can create a folder of the model and just put there the model and index file or you can just throw the model in the logs folder itself
simple ore
This GeForce RTX 5090 size looks thinner than RTX 4090. <:Baffled:11593972868335...
it uses two passthru fans while the card itself is super tiny
inland rover
Hello, i'm trying to train a model with RVC v2 Disconnected on Colab, i followed every instructions and started the training, it's still at 0% at Epoch 318 i only wanted to know if this was normal or not ?
pearl cove
https://docs.aihub.gg/rvc/local/mainline/
It’s still the same
odd shale
It’s still the same
Then i don't really know what could cause that "cutoff" on your audio when inferencing with models.
Sorry.
pearl cove
I can’t figure it out too cuz it was working without problems before
coral vault
ok
inland rover
:(
hallow thistle
Then i don't really know what could cause that "cutoff" on your audio when infer...
The "cutting off" audio can occur with a badly-trained voice model. That's what I know.
knotty moth
It’s still the same
I'm not sure what "cutting off" you mean, could you show the spectrogram using spek or audacity?
pearl cove
Uh lemme download
ember wadi
@inland rovercheck the logs folder, or the asset/weight folder, sometimes percentage dosnt' update much, but you can still get an out put
odd shale
The "cutting off" audio can occur with a badly-trained voice model. That's what ...
Hm, maybe you're right.
pearl cove
I'm not sure what "cutting off" you mean, could you show the spectrogram using s...
Like that?
pearl cove
The "cutting off" audio can occur with a badly-trained voice model. That's what ...
Yuh but the model worked perfectly before used it couple of times now it sounds ass and choppy and cuts out and robotic it was perfect before tho
ember wadi
Yuh but the model worked perfectly before used it couple of times now it sounds ...
a wild guesse on hop_length
knotty moth
which is the original one and which is the inferred one?
knotty moth
well didnt see the "cut off" part
pearl cove
Well it’s like getting quieter at parts where the part kinda cuts off
I don’t see it on the track too it’s the same size but still gets quieter
knotty moth
you tried playback in audacity? sounds like something wrong with the audio device itself
pearl cove
Yea it js sounds like it sounds on fl the same result
brittle wing
The -colab command doesn't work, what's happening
ember wadi
The -colab command doesn't work, what's happening
I've tried all the RVC colab link by bot, none of them works, i guesse Appio still working
brittle wing
I've tried all the RVC colab link by bot, none of them works, i guesse Appio sti...
Wdym
ember wadi
colab updated python to --version .11
so most of the dependency are un compatible
imcompatible
brittle wing
They work?
They're here but why does the command not work are they updating something
ember wadi
maybe im not using the right version? i don't know
you can downgrade colab to python3.10, you will need to add version infront all the cli command, including !pip...
it's not a solution, but, it will patch the problem for now
coral vault
how do i change the allowed_paths parameter of launch when using applio rvc
brittle wing
it's not a solution, but, it will patch the problem for now
The bot command doesn't work
-guides
How do I report the problem to an admin/staff?
There's a guides section but how about the Colabs?
steel forge
How do I report the problem to an admin/staff?
That bot is down rn
Idk when or if it'll come back
We do have a modmail bot if you need to reach out, @vital hedge
@azure marsh I miss you
#📰│dev-updates message
^^ working colabs
brittle wing
That bot is down rn
So it's malfunctioning
brittle wing
Ok
low shard
https://discord.com/channels/1159260121998827560/1159380240271953940/13293800726...
I hope users will still check that message even if they have to scroll a bit
neon barn
Hi, could someone please give me the download link for the voice changing app?
low shard
Hi, could someone please give me the download link for the voice changing app?
realtime voice changer for calls? wrong channel, RVC stands for Retrieval-based-Voice-Conversion, which is speech to speech, but you will need Wokada that uses it for realtime
tell ur pc gpu in #🔍│help-w-okada
neon barn
Thanks. Sorry for the mistake.
low shard
Thanks. Sorry for the mistake.
it's fine
brittle wing
I trained a 6:43 model with batch size 4 yet it doesn't sound like the person
Should I increase the batch size
I used ov2 super tho
crude flame
you should get a bigger dataset
brittle wing
you should get a bigger dataset
But the data of that person is quite limited, I have only one song where he doesn't use so much autotune
https://www.bandlab.com/post/6f159c132108e9119f2a00155d77cb96_5596e3ebe68ee911abc428187831ee7a
brittle wing
you should get a bigger dataset
Also he has a lot videos where he talks
Uh at least 30 mins
crude flame
then use those
brittle wing
then use those
But they have autotune & one of the songs is 128 kbps mp3 & everything else is m4a quality
crude flame
why does he have autotune on talking audio
brittle wing
then use those
But it's gonna be a lot of talking w 2 minutes Rap honestly
crude flame
But it's gonna be a lot of talking w 2 minutes Rap honestly
thats fine
brittle wing
thats fine
Only the elevate song is the cleanest sounding
brittle wing
thats fine
Will it still emulate his rapping
crude flame
Will it still emulate his rapping
it will sound like him but to emulate his style that would depend on the inference audio
brittle wing
Also will making a model of a Korean girl that sings in English who had a racist past towards black ppl expel me from the server
Should I post it or keep it for myself.
I decided to ask if that matters
brittle wing
it will sound like him but to emulate his style that would depend on the inferen...
So the speech wouldn't be such a problem uh nice
crude flame
Should I post it or keep it for myself.
i think you should keep it to yourself
brittle wing
Cause I'm gonna need his voice to generate some ai songs and stuff
I understand I won't post it there
The vids of him talking resemble his rapping voice
brittle wing
i think you should keep it to yourself
Okay I'll only share it on HF for backup
craggy vector
hello, guys, i have a problem with voice changer, where can i share my problem and ask for help?
low shard
hello, guys, i have a problem with voice changer, where can i share my problem a...
Realtime voice changer for calls? Wrong channel, Rvc stands for Retrieval-based-Voice-Conversion
You need to tell me ur PC GPU in #🔍│help-w-okada
brittle wing
i think you should keep it to yourself
Will there be a problem if I post it?
I know that if I post it admins can delete it if they don't approve
loud talon
ruby lava
how do i fix this error? 'NoneType' object has no attribute 'setdefault'
knotty moth
Should I increase the batch size
I used ov2 super tho
use the default pretrain if not sure
knotty moth
But they have autotune & one of the songs is 128 kbps mp3 & everything else is m...
128 kbps mp3 sources are worse than youtube streams to make as dataset
you can try apollo universal enhancer but don't expect uncompromised quality
brittle wing
128 kbps mp3 sources are worse than youtube streams to make as dataset
you can t...
I thought of this but it's gonna waste my mvsep credits
Well I download in opus when it comes to getting best YouTube quality
brittle wing
128 kbps mp3 sources are worse than youtube streams to make as dataset
you can t...
It's not even YouTube streams they're videos from 2019 of a man recording himself speaking on his phone...
knotty moth
I thought of this but it's gonna waste my mvsep credits
there is apollo colab
brittle wing
there is apollo colab
Actually I'm gonna reprocess the song only also the audio is short.
knotty moth
I thought of this but it's gonna waste my mvsep credits
btw it's just for convenience stuffs, you can still access the models with the free account
brittle wing
btw it's just for convenience stuffs, you can still access the models with the f...
I know but I had to use credits for the speech
Cause it's more than 10 mins
knotty moth
I know but I had to use credits for the speech
- apollo colab has universal & speech enhancer options
- you can always split the audio manually without cutting some sentences
hidden linden
For voice changer work, i need python and what else is the second one?
brittle wing
- apollo colab has universal & speech enhancer options
- you can always split th...
My bad for not thinking about that earlier
Wait can I enhance the speech after denoising
knotty moth
Wait can I enhance the speech after denoising
enhancing adds noise
brittle wing
enhancing adds noise
Ouch I know
sage trench
crude root
im on mobile and when i try to make a song cover, there is no option to add a youtube link for the song. how do i do it?
hallow thistle
crude root
wow thanks for responding so quick. do u have any idea when it will be working again?
hallow thistle
wow thanks for responding so quick. do u have any idea when it will be working a...
Team Weights may soon update the news in their Discord server regarding the issue. Right now, all you do is wait.
knotty moth
im on mobile and when i try to make a song cover, there is no option to add a yo...
as a workaround, you can download the video using https://github.com/yt-dlp
jade pine
where do i get RVC?
low shard
where do i get RVC?
What's ur PC GPU and what are u looking to do
jade pine
What's ur PC GPU and what are u looking to do
3060ti and just trying to mess around with voices in games, I want to make some videos just messing with different models
low shard
3060ti and just trying to mess around with voices in games, I want to make some ...
So, use a realtime voice changer for videogames and record it right?
jade pine
Yea
lunar pebble
hello
low shard
Yea
then you don’t need RVC, it stands for Retrieval-based-Voice-Conversion, you need Wokada that uses RVC in realtime
will send u the link in #🔍│help-w-okada
jade pine
Alright thank you
lunar pebble
can i use AI here
formal wind
Welp yall mods know what to do
pearl cove
Does anyone know why it’s getting quiet at some parts
midnight fern
guys is there any websites that rap your lirycs??
simple ore
suno?
polar tendon
Does anyone know why it’s getting quiet at some parts
Might be due to the instrumental being very loud at that time, louder than the vocals
pearl cove
Might be due to the instrumental being very loud at that time, louder than the v...
U could be onto something but I don’t think it’s bcuz of that, forgot the raw vocals rn need too check em again based on what u said
pearl cove
Might be due to the instrumental being very loud at that time, louder than the v...
Nah it’s clear bro
spare junco
guys where i can download the voice changer?
silent stratus
Does anyone know why it’s getting quiet at some parts
prolly bc u were talking quieter, use one of my trav models too and see if it fixes it
actually it might be ur preset
remove the room reverb and it would be pretty good
silent stratus
guys where i can download the voice changer?
orchid flint
Anyone know why this might be overtraining so early? Its a 28 minute dataset with a batch size of 8.
carmine siren
Anyone know why this might be overtraining so early? Its a 28 minute dataset wit...
There might be having some issues with yours Dataset,
orchid flint
Not sure what it would be. Its ripped directly from a game
knotty moth
Not sure what it would be. Its ripped directly from a game
some of them may be in lossy mp3, ogg, acc compression
also it mostly still needs cleaning process
orchid flint
They were exported in WAV
carmine siren
some of them may be in lossy mp3, ogg, acc compression
also it mostly still need...
There might be not having clear dataset that should required be for better training
orchid flint
I can send you one of the files though DM if you want
knotty moth
They were exported in WAV
the format isn't relevant, but it may come from such lossy compression, as seen in the spectrogram
orchid flint
simple ore
Anyone know why this might be overtraining so early? Its a 28 minute dataset wit...
it is a pretty flat charts - 36,2.. 37.4 is not much variation to be concerned about.... but also you're looking at the old chart with random batch loss value
pearl cove
prolly bc u were talking quieter, use one of my trav models too and see if it fi...
Ngl I think this one uses yours idk tho imma check later 🤞🏽
open stag
- colab
keen coyote
which versions the newest i think mines outdated
low shard
which versions the newest i think mines outdated
no need to ask the same thing every channel, u got already replied to in #🔍│help-w-okada
pastel oak
When people join a hub which implies the existence of multiple ai softwares in different area usage and its always "my thing doesnt work"
knotty moth
When people join a hub which implies the existence of multiple ai softwares in d...
too bad not most ppl in such discord servers know how to ask in typical support forums
patent trellisBOT
@vukannn banned
Reason: Our algorithms suspect your account may be compromised due to unusual behaviour, such as spamming or posting inappropriate content. Please secure your account and appeal to restore access.
Duration: Permanent
hallow thistle
That's unfortunate. You should've never click on those links, and make sure your account has 2FA mode on.
low shard
When people join a hub which implies the existence of multiple ai softwares in d...
real
acoustic coral
My Okada website won't work, any suggestions?
craggy saffron
LATINA E GIRL VOICE!!1!1111
knotty moth
craggy saffron
damn
low shard
My Okada website won't work, any suggestions?
replied already in #🔍│help-w-okada , and this is the wrong channel
brittle wing
Hey yo I'm trying to get
Sorry one sec
I'm trying to create an ai cover of a song that literally just changes a few lyrics, same singer same everything just a few lyrics changed, anyone know how I could do that?
So i guess it's not really a cover more like the lyrics just changed
I have isolated vocals of the audio I just am trying to figure out how to change the lyrics
rapid zinc
guys, I wanted to make ai cover with RVC-HF, but after pressing “Duplicate this space” I keep getting this error, can you guys help me fix this please?
low shard
I'm trying to create an ai cover of a song that literally just changes a few lyr...
you have to sing the modified lyrics
brittle wing
you have to sing the modified lyrics
True
low shard
guys, I wanted to make ai cover with RVC-HF, but after pressing “Duplicate this ...
That one is made specifically for ZeroGPU, which is a thing you get if you paid for PRO
rapid zinc
That one is made specifically for ZeroGPU, which is a thing you get if you paid ...
So are there any tools that I can use for free to make Ai covers? For example, if I use the WebGui version or the colab version of this tool, do I still have to pay?
low shard
So are there any tools that I can use for free to make Ai covers? For example, i...
You can just use that HF (HuggingFace) Space without cloning it, you shouldn't clone ZeroGPU ones
Also, no you don't have to pay on colab
-hf
low shard
there's AICoverGen
there's also https://weights.gg
did you try checking your pc gpu before to see if its good enough tho?
bc all of those are on cloud (remote good pc), for people who got a bad pc
karmic oliveBOT
<@1332283502082920508> Saiki Kusuo voice changer
warm isle
@karmic olive help
karmic oliveBOT
<@1332283502082920508> help
karmic oliveBOT
<@1332283502082920508> rcv
hallow thistle
Bro just left the server because he didn't know how to use Automaze. 
brittle wing
@proven hill original pretrain isn't the best
brittle wing
Hello does becruily's vocal model really implements noise in lead vocals/Acapellas?
runic trellis
I think APPLIO is not working, when executed it does not deliver any results, does this happen to anyone else?
marble forge
is there any ai software that can turn a song into something that sounds like one of those wind up toys from the 1900's (like a snowglobe with music) or is there any way to do this? This is sometimes also called a music box.
low shard
i need hel with kaggle
!howtoask
low shard
Elaborate more
low shard
I think APPLIO is not working, when executed it does not deliver any results, do...
Elaborate more on what u are doing, using, and what's ur PC GPU
brittle wing
mantrax died 
brittle wing
mantrax died <:trolley:1159468147133395025>
he's on the sewer, eating melted cheese /j
little ore
how do i use jp huber in rvc realtime voice changer?
karmic oliveBOT
Interaction has expired, use the command again for a new interaction.
low shard
how do i use jp huber in rvc realtime voice changer?
this is the wrong channel, Wokada uses RVC for realtime, tell me your pc gpu in #🔍│help-w-okada
brittle wing
how do i make the applio colab notebook download models from weights.gg
brittle wing
how do i make the applio colab notebook download models from weights.gg
applio can't download models from weights.gg
bruh
what do i do then
you can manually download the zip file of the weights model
extract it in a folder
go to applio and upload the pth and index files
so just unnecessarily making it harder
well
i got no other choice
brittle wing
so just unnecessarily making it harder
sadly yes, applio can only download models from huggingface
they gotta implement weights support
true
low shard
so just unnecessarily making it harder
it's not their fault,
Weights.gg needs login to download them (which is made so other users won't just make their own site and re upload everything somewhere else)
and iirc there isn't weights.gg api support yet
flint solar
sadly yes, applio can only download models from huggingface
It can download from weights
hexed vault
how to uninstall
low shard
how to uninstall
how to uninstall what? rvc? u should be fine by just deleting the folder
tame mica
bro dipped
brittle wing
how to uninstall
!howtoask
wanton frost
hi people, i was using rvcDisconnected v2 by "KitLemonFoot". and it worked well for me yesterday. but today i faced this issue where the loading stops when it reaches "clone repositories". that's the related screenshot for the issue. please don't forget to @ me when answering! thank you, and sorry if i bother anyone :>
- i tried shutting the tab and opening it again, but to no avail
brittle wing
hi people, i was using rvcDisconnected v2 by "KitLemonFoot". and it worked well ...
seems that google colab updated its python
i had a similar error with nick's gpt sovits webui colab
lavish lintelBOT
wanton frost
i had a similar error with nick's gpt sovits webui colab
ohh, so should it resolve itself soon?
brittle wing
if lemonfoot tweaks the colab, maybe it could be fixed
while you wait, check out these colabs:
-colab
wanton frost
if lemonfoot tweaks the colab, maybe it could be fixed
thank you so muchh, coul
brittle wing
you're welcome
wanton frost
oh i was just about to ask
brittle wing
try using rvc mainline or applio colabs
wanton frost
ilysm thankss<333
brittle wing
ilysm thankss<333
love you too, you're welcome
wanton frost
try using rvc mainline or applio colabs
i'll try!
low shard
hi people, i was using rvcDisconnected v2 by "KitLemonFoot". and it worked well ...
Damn it's broken again??
Lemme check it out
low shard
if lemonfoot tweaks the colab, maybe it could be fixed
It actually was said it's been fixed in #📰│dev-updates
wheat plover
I'm confused about Formant Shifting, if I have a male model, and a female audio, I use f2m or m2f?
simple ore
I'm confused about Formant Shifting, if I have a male model, and a female audio,...
it applies to input audio
so you want to shift the female voice to the male voice range
low shard
@wanton frost @brittle wing Alex deleted his mangio rvc tweaks fork, meaning that rvc disconnected is broken again since it's based on that, I wouldn't believe that it's gonna come back after this 
@rigid mango u should know this btw
brittle wing
<@1165691437543850065> <@456226577798135808> Alex deleted his mangio rvc tweaks ...
aw shucks
i thought that it was because the notebook wasn't adapted to colab's python update
wanton frost
<@1165691437543850065> <@456226577798135808> Alex deleted his mangio rvc tweaks ...
OMG NOOOOOOOO rip that was my comfort way of training
low shard
OMG NOOOOOOOO rip that was my comfort way of training
You gotta adapt to newer ones
Maybe it will come back, Maybe not
Who knows
Because the dev is busy, and has no control over the GitHub fork since it was of someone else
And Alex deleted it
brittle wing
<@81458568141733888> u should know this btw
all the rvc disconnected model makers now have to adapt to mainline or applio
low shard
So unless he bases his entire colab around another rvc, which takes time, then rip
brittle wing
rip rvc disconnected
rigid mango
😑
I'll switch it back to the original mangio in a bit
It'll be slower but it will be functional
wanton frost
oof thank you guys still for informing on that. will miss it for sure
also mainline vs applio, which one do y'all recommend?
analog obsidian
oof thank you guys still for informing on that. will miss it for sure
also main...
applio
mainline is too old and outdated
wanton frost
thanksss<33
low shard
I'll switch it back to the original mangio in a bit
It'll be slower but it will ...
Alright, pls let me know when this is done
wanton frost
sorry for questions one after another but which pretrain do you guys recommend for 48k samplerate audios - for training?
low shard
sorry for questions one after another but which pretrain do you guys recommend f...
Depends on your dataset language and length
Usually the OG one is suggested for English ones
wanton frost
i'm on mobile and i've seen the latest ones mostly made for 32k and 40k, so it seems like the ways to change that samplerate as you guys do in computers is not as easy for me
// my language will be turkish, a language where every letter is read as it is.
example: when you use "c" it's not ever a "a, b, c(see)" sound or a "c(k)ook" sound
and my datasets are usually around 3-5 mins
analog obsidian
original pretrain waa trained in english only, you can train any language using it but in rare occasions it doesn’t sound good in some languages
rigid mango
Alright, pls let me know when this is done
switched it back over
low shard
switched it back over
Alright, thx for the fast reply
Will the speed be much slower?
Dunno the differences between normal and Alex mangios
teal cradle
so like
wanton frost
original pretrain waa trained in english only, you can train any language using ...
well yeah then thank you!
rigid mango
not sure, but my guess is probably a fair amount
teal cradle
is rvc disconnected becoming connected sorry im stupid
analog obsidian
well yeah then thank you!
expect some slight mispronunciations
teal cradle
cant someone just reupload the rvc stuff
low shard
cant someone just reupload the rvc stuff
Nope the Alex mangios fork is deleted, now it will use original mangio which is slower
alpine lotus
is rvc disconnected becoming connected sorry im stupid
oh hi sads
low shard
not sure, but my guess is probably a fair amount
rip
analog obsidian
just a reminder mangio fork havent been updated in almost 2 years
low shard
Yeah
teal cradle
dude were so cooked 🥀
low shard
Not to sound rude, but I don't get why people don't just use Applio that gets more updated
analog obsidian
its even behind mainline in updates
low shard
Sorry if it sounds any rude
alpine lotus
is rvc disconnected becoming connected sorry im stupid
also the creator behind rvc disconnected is no longer maintaining the colab bc he isn't really interested in rvc anymore iirc
crude flame
Not to sound rude, but I don't get why people don't just use Applio that gets mo...
RVC disconnected is faster !!!!!!!!!!!!!111
teal cradle
RVC disconnected is faster !!!!!!!!!!!!!111
and a lot easier
crude flame
and a lot easier
💀
low shard
and a lot easier
His message was meant to be ironic
analog obsidian
crude flame
I was being sarcastic
teal cradle
oh
low shard
RVC disconnected is faster !!!!!!!!!!!!!111
I don't wanna offend kit ofc, but yeah it's slower than applio and mainline now
alpine lotus
both rvc disconnected and applio have almost the same training speed.
alpine lotus
iirc
rigid mango
That's kind of why I don't really want to keep updating disconnected
So much of the notebook is hardwired into mangio to retain backwards compatibility that it'd basically require a full rewrite to switch it over to something like mainline or applio
analog obsidian
Applio training is hella fast rn
low shard
both rvc disconnected and applio have almost the same training speed.
That changed for sure now, applio is way more optimized and disconnected uses original mangio now
teal cradle
i just like the original pretrain or whatever
and how i can use google drive with it
low shard
That's kind of why I don't really want to keep updating disconnected
So much of ...
Welp, what do you think is right to do?
You're the creator afterall
wanton frost
Not to sound rude, but I don't get why people don't just use Applio that gets mo...
we are js stupid and are more relying on some "now you do this, now you do that" tutorials around youtube which explain everything to us as we're some kindergarteners(and it seems correct)
alpine lotus
That's kind of why I don't really want to keep updating disconnected
So much of ...
btw any updates on the vtuber rvc voice model sheet?
crude flame
i just like the original pretrain or whatever
The applio notebook can do both of those
low shard
I honestly would think it's best for people to adapt to the newer and faster forks, I guess people use it because they just don't wanna adapt
wanton frost
I honestly would think it's best for people to adapt to the newer and faster for...
well basically that lmao
low shard
we are js stupid and are more relying on some "now you do this, now you do that"...
YouTube tuts are severly outdated, you're just gonna use a thing that is slower and less updated, we have docs http://docs.aihub.gg/ which explains users how to use newer things like applio
alpine lotus
I honestly would think it's best for people to adapt to the newer and faster for...
i mean i have adapted to applio and now i got a bit used to it, i have some complications with it but i'm getting there.
low shard
well basically that lmao
Yeah but at the cost of it being even more slower now
rigid mango
Welp, what do you think is right to do?
I'd really like it if someone out there could take the Colab-centric design philosophy of Disconnected and upgrade it to a modern version of RVC, but that person is not gonna be me unfortunately
wanton frost
YouTube tuts are severly outdated, you're just gonna use a thing that is slower ...
oh well i didn't know that even existed either lol thank youu<3
alpine lotus
I'd really like it if someone out there could take the Colab-centric design phil...
rejekts you know what to do 
low shard
I'd really like it if someone out there could take the Colab-centric design phil...
With that philosophy you mean being no UI? Because applio no ui exists too even if less used than the UI one
Since UI is generally easier
analog obsidian
With the correct slicing and using applio, you can have a model sounding good in 100 ish epochs
While in mainline it would take 200-300 or even more
Sure its boring to learn how to use the thing but its worth it
low shard
rejekts you know what to do <:trol:1159397152183824405>
Rejekts is not active either, I'm not even sure if the newer easy GUI works
Did anyone even test it recently
I might have to soon
rigid mango
With that philosophy you mean being no UI? Because applio no ui exists too even ...
No UI, and everything done in Colab's interface using markdown instead of gradio/ngrok
I honestly expected Applio No UI to kill Disconnected at some point but well, here we are
alpine lotus
No UI, and everything done in Colab's interface using markdown instead of gradio...
btw any updates on a vtuber rvc voice model hf space and sheet?
bc i have made some more vtuber models after it got updated
rigid mango
btw any updates on a vtuber rvc voice model hf space and sheet?
none atm
I know you've made some more stuff but haven't gotten around to it
I've waned off of vtuber voice AI pretty hard, since there's not really anyone doing it anymore
low shard
No UI, and everything done in Colab's interface using markdown instead of gradio...
I mean you're describing what applio no ui does, it's just not used bc there's no guides for them
Do u really want it to "kill" ur notebook tho
rigid mango
I honestly don't mind at this point
low shard
I honestly don't mind at this point
Wanna talk in #💬│engineer-chat a bit?
rigid mango
Not really
low shard
Not really
Ah, I was going to ask if you were going to keep supporting the colab in the future
Anyways
It's listed as fixed colab in #📰│dev-updates
I swear mangio Is gonna delete the fork TMR or smt atp 😭
This gotta be crazy unluckiness that Alex randomly deleted his fork
lavish lintelBOT
brittle wing
now people yes or yes have to switch to applio, the ui or no ui notebook
mangio rvc and mainline aren't very maintained at all (basically they're outdated)
low shard
mangio rvc and mainline aren't very maintained at all (basically they're outdate...
Mainline is still maintained iirc, mangio is archived since more than a year
brittle wing
Mainline is still maintained iirc, mangio is archived since more than a year
but that doesnt mean that it isnt old/outdated
alpine lotus
Mainline is still maintained iirc, mangio is archived since more than a year
oh hey that reminded me, remember when kalomaze was still here and he announced that they'll be making mangonada back in 2023?
brittle wing
oh hey that reminded me, remember when kalomaze was still here and he announced ...
mangonada?
brittle wing
to make voice models sound better?
if it's that, it may help model makers with voice models
silent stratus
would a gtx1650 work with codenames fork or should I figure out a way to port it to colab or kagglr
slender osprey
Is it possible to use some version or I don't know is there any way to use vcclient without cuda?
brittle wing
RVC Disconnected is really outdated fr
https://colab.research.google.com/github/IAHispano/Applio/blob/main/assets/Applio_NoUI.ipynb#scrollTo=Ncjk1wkrNpuo
How do I add a pretrain here for training???
How do I resume training on there tho?