Im getting this error when doing the last cell:
#so-vits svc error
thorny skiff
using less outdated software usually helps
uncut inlet
Im getting this error when doing the last cell:
-svc
tardy plover
digital tusk
Im getting this error when doing the last cell:
Use rvc, svc is from dinosaur ages
umbral beacon
Also what's your PC GPU btw
gtx 1660 Super
umbral beacon
Be sure to not use yt tuts
what should I use instead
astral ivy
gtx 1660 Super
that would be good enough for model inference (use models) but eh not that much for training (making models) locally
astral ivy
what should I use instead
what are you looking to do?
astral ivy
like putting the models over vocals
that's called inference
Your Nvidia GPU is good enough to do inference (use models) locally (on ur pc), not the best to train (make models) even if still possible
You can:
- Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
- Cloud (remote good pc, easier and faster than ur PC but it's limited):
- Ilaria RVC Zero: fastest and simplest that you can get for free
- Weights.com: Partnered with AI Hub, lets u do them easily but u may be in a queue
- Applio UI Colab: max 4 hours daily, not granted, of GPU (needs workaround)
- RVC-AI-Cover-Maker-U Colab: Automatically separates the vocals and instrumentals, converts the voice and mix all together back
Easiest possible (automatically separates vocals & instrumentals) : weights.gg & rvc-ai-cover-maker-ui
easiest cloud: Ilaria rvc zero
easiest local: Applio
umbral beacon
thank you
soft orchid
Im getting this error when doing the last cell:
I still use it. Don't listen to them. Rvc actually has a better quality aspect to the output. I also love sovits because each output is never the same as the other nor to the og reference. Helps with doubling effect.
uncut inlet
I still use it. Don't listen to them. Rvc actually has a better quality aspect ...
please give me some cases where sovits svc is better than rvc
astral ivy
Also not true u can use a notebook
the project is abandoned and even older than so vits svc
so vits svc is harder than rvc and lower quality, no reason to still use it
soft orchid
Not true. Quality wise so-vits-svc has a better depth.
pearl temple
uncut inlet
Not true. Quality wise so-vits-svc has a better depth.
instead of saying such BS, try training the same dataset using sovits svc and rvc v2 then show the result samples
soft orchid
Instead of having your head up ur ass why don't you?
pearl temple
Are you perhaps using phone speakers to listen to the audio? RVC is so much better
soft orchid
pearl temple
https://youtu.be/59A5uxIKw1s?si=AgR_QnQFSrYx-mEF
💀
That is not a valid source
1 year old video and the model was not trained properly
Probably also not even cleaned properly
soft orchid
There is some interesting information. I am not saying sovits is better, but i have a use for it. Dealing with vocals that apply with mixing and the effect im aiming twords
For now that is.
That video is a few months old pal what are you talking about
unkempt trellis
https://youtu.be/59A5uxIKw1s?si=AgR_QnQFSrYx-mEF
thats a rvc v1 model
not even v2
pearl temple
And it's v1 💀
pearl temple
Why do you hold yourself back so much
uncut inlet
thats a rvc v1 model
iirc rvc v2 has already been there since I joined ai hub just before the takedown incident in 2023
though I got some v1 models at that time
uncut inlet
That video is a few months old pal what are you talking about
good point, time traveler!
opaque spindle
And it's v1 💀
rvc v1 is much better ive been telling u this for months, v2 has less accuracy and has more robotic sounds and noise than v1
i sent proof of these claims with samples from the same dataset trained on 40hz and 48hz on both rvc v1 and v2 and v1 outperformed
some people made the excuse that rvc v1 must have "better pretrains for rvc v1 for tupacs style" so i even tested other models to break that theory and every model i trained still came out more realistic than v2
and ive been here since the beginning and tacotron 2 etc ive tested these things over and over compared everything talknet so-vits Applio rvc nothing compares to rvc v1 yet
astral ivy
@umbral beacon , so how's it going? is this all solved?
deft notch
i sent proof of these claims with samples from the same dataset trained on 40hz ...
got one sample of the difference?
astral ivy
Let's not turn this into an heated argument, this is an help channel, and we provide help for RVC v2, which is the latest and best quality as of the moment, there's no point into using outdated architectures
deft notch
i sent proof of these claims with samples from the same dataset trained on 40hz ...
just share in ai support channel if u want to
umbral beacon
<@539238379347509289> , so how's it going? is this all solved?
Yea
astral ivy
Yea
I'm guessing I can mark this as solved then
if you need any other help, please make a new post
umbral beacon
ok
opaque spindle
got one sample of the difference?
ive already done this months ago brother i cant be bothered to go through all this proof thing again
opaque spindle
# Let's not turn this into an heated argument, this is an help channel, and we p...
just saying it in a friendly way my friend no arguments my way
astral ivy
just saying it in a friendly way my friend no arguments my way
please yall use #🧬│ai-chat or #🔊│ai-development for such things, this was just an help forum
opaque spindle
gotcha!
soft orchid
please yall use <#1159289354439626772> or <#1159290193619189821> for such things...
Ok so since this is a help forum. Can you explain why is makaveli getting greater results in his research and tests? What is the difference between applio and rvc "outdated" architecture besides that its no longer being developed? My guess is maybe applio trains differently? Can you point me into the right direction of why? Is it maybe rvc does better with bigger datasets apposed to applio's Is efficiency with smaller datasets? 🤔 I have made decent ones with 5 min datasets. Im just curious. Maybe makaveli is onto something?