#Anything working with mac os ?
1 messages · Page 1 of 1 (latest)
Ayo? @wicked vigil level 1 !!! 
The realtime voice changer should work on an Apple Silicon Mac (M1 and above)
EOL - No further Updates
Github - Blanc-dot
Discord - Blancdot
Other Links
Antasma's Local Error Fixes
Antasma's Colab guide
Sushi's useful Links - You need to follow this if you are on an AMD or INTEL ARC graphics card
Frequently Asked Questions
Jump To The Important Sections
Other Links
Downloa...
Anything else is kinda meh on Mac support
And Colabs normally have 4 hours of runtime for free so
what happens after the 4hours runtime ? It has a "cooldown" ? or you just cannot use it anymore ?
yeah it disconnects and you need to wait 24 hours
Okay, thank you ! Is there any thing to bypass it ? VPN or anything ?
Also, i tried for the first time making an AI cover, i downloaded a well built model here, uploaded a good quality song, the voice are good inded but there are some really weird artifacts in the vocal and with some notes, also the background vocals going absolutely wild. It's the same with 2 different songs. Is the issue from the model i downloaded or from the program i'm using, which means i should use another ?
nothing sadly
unless u pay $10
you can use another acc to get more time but... eh
eh like it doesn't work or what's the issue ?
background vocals make it go wild yeah, you need to use the main vocals only
how do i use only main vocals ?
ya need to filter em out, the guide should explain that
i'd be willing to pay, but isn't that more or less "illegal" because i guess ai covers isn't really legal, is it ?
where is the guide please dude ? I'm all new into this. I also have some songs already separated (vocal and instrumental) but it seems to me that i cannot upload these files, because the program makes the separation on its own...
unless you're distributing it online somewhere... not really illegal tbf 
oh you're using aicovergen
just use another colab, like ilaria rvc
https://docs.aihub.wtf/rvc/cloud/ilaria-rvc/ for the tutorial
Last update: Mar 8, 2024
that site also has ai cover tutorials
;)
yes aicovergen ! is it the "worst" ?
Can you please explain to me, with the same model, the same song, can several different programmes provide different results? In other words, is there an algorithm specific to each programme?
Can Google ban an account that uses Colab to do music AI?
Ayo? @wicked vigil level 2 !!! 
no, there isn't a specific algorithm to each programme
it's just that some forks (variations of RVC stuff) have different ways of doing stuff. AICoverGen already does the separation part for you, although it fails at times.
Google won't really ban if you do Music AI, it'll ban if you break their ToS ig
Thank you ! Did you guys, i mean staff of this discord server, ever thought about doing a huge backup of all the tutorials/models files so if somehow the discord gets strike you'll still have the resources ?
One more question please, i tried downloading a model, but when clicking the link i get an error message saying "invalid password" something like that. What can i do ? Nothing i guess ?
weights.gg is the backup of everything, we did this cuz we already got banned before
use a different model in that case... most likely privated model
Okay ! One thing i didn't like with AiCoverGen is the background vocals going completely crazy as well as some really weird artifacts in some notes, and also the audio levels of the main vocals getting too much loud compared to the original sound, things like that. That's why i'd like to upload my own separated tracks.
yeah that's why you should use Ilaria RVC then
already sent ya the guide in here ;)
Is there anything really good, reliable, for ai covers, that can run on cloud or mac whatever, even if i have to pay for it ? The google colab is good enough, but i dont really like the runtime limitation, i'd like something that i could use without being really limited.
What's happening here, i closed my previous aicovergen colab tab, tried opening the one you sent me, but google keeps telling me i have another tab still runing, but i closed it... ? Is that even normal ?
Hm... you could try running Mainline locally: https://docs.aihub.wtf/en/rvc/local/mainline
But not sure if it'd work well tbf 
404 error xD
Oh yeah that's normal
woops i forgot that the link's old lol
What should i do then ? 😅 i closed the tab, but it keeps saying i've got an active tab.
xD still 404
wait im not doing this right
haha okay
yeah you need to manage the runtime, top right corner has a lil dropdown button alongside some little performance graphs, click that and manage runtime
then just delete the one that's running
https://docs.aihub.wtf/rvc/local/mainline/ if u want the local thing, idk if it'll work well tbh but ye
Last update: Mar 8, 2024
NOW IT SHOULD WORK haha
it works now ! haha thank you
Please let me ask you another question, now that i can upload my own separated tracks, i do the separation with "ultimate vocal remover" but do i really absolutely need to : separate vocal from instrumental => put the vocal only track in ultimave vocal remover and remove the reverb and only then upload to the AI cover program ?
Is that absolutely mandatory ?
you dont need to remove reverb unless it's that bad
also, use stuff like UVR BVE to get the main vocals, the main vocals part is mandatory
what is UVR "be" ? i know about the "basic" uvr but whats "be" ?
Ayo? @wicked vigil level 3 !!! 
the BVE is a model to get the backing vocals and the main (lead) vocals
where can i find that please ?
should be in the VR Arch models
else just head to the download center and download it
Found it ! Do you know any sort of documentation that explains the differences between all the different models of UVR and their purposes ?
Amaaazing ! And from your personal recommandation, do you know any UVR model that works good to extract adlibs from an acapella ? example : i have a vocal only tack and the singer is basically sing, but in background there are also adlibs going at the same time than the lead vocal, is there a mode more or less capable of separating again lead vocal and the adlibs ?
i guess BVE could do so
but not entirely sure... 
thank youu
yw :)
what is the "pitch extraction algorithm" do i have to play with it ?
that's how RVC knows what pitch to use
normally id just use rvmpe
but if u wanna see if something else can bring a better result you can always use mangio-crepe
okay i guess it's an important setting, right ?
I just tried with the forked you sent me, and i'm really getting some really weird results, i don't understand what am i doing wrong. Maybe because the vocal only track has some harmonies/subtle choirs in it ? If yes, how do i remove them with uvr ?
Oh maybe it's because of the harmonies yeah
have you already used BVE on it? if so I think you should try other Karaoke separation methods, there's 5HP and 6HP karaoke models on VR Arch, see if that works for that
Dude so disappointed with the results i'm getting xD I was really expecting better because of all the AI covers i got to listen to, i can never seem to get even close to any of them. I think my vocal tracks are fine, where did i go wrong then ? Would you please be able to help in dm ?
Don't think you're going wrong... maybe it's just the song sometimes
Also I don't really like going in DMs 😭 sorry
when the title of a model says "rvmpe" does it mean that we should not use any other pitch extraction algorithm when the fork offers "mangio crepe" for example ?
np
you can use the others too
rvmpe was just used on the training
thank you, you rock
you too ;)
Hey dude ! I've been practicing a little more with AI, and indeed i came to the conclusion (lol) that if i want a good result, i absolutely need a really really clean accapella track. That's the tricky part. I used some random songs i had, passed them through UVR i tried 3 or 4 models, but absolutely none of them allowed me to have a good result after processing it in the fork. But i guess there is no secret magic here, i just need to know how should i use properly UVR. The only songs that gave me a good result were the songs that i could find their official accapellas (which means the accapella is ultra clean). Do you have any tips please in order to have the best accapella possible with UVR ?
Sadly I'm not so sure on how to get the best results on UVR, I don't really use it tbf (I use MVSep since I don't really wanna download UVR as of rn)
But I think you could try using this sequence of models to try and get the best lead vocals:
MDX23C or BSRoformer (both of which can be downloaded on the Download Center) for all the vocals
UVR BVE or 5/6HP for the leads
De-Reverb and/or De-Echo to remove reverb stuff
Ig those are the best ones... you can use them in sequence from all vocals to reverb removal
If my understanding is correct, i should follow this path :
Normal song to UVR with MDX23C to separe instrumental and vocals, then take the vocals and pass them through UVR BVE and if needed, pass them now through De-Reverb ? Is that right, right order ?
It would be awesome if you could tell me how to install Applio on macos? I'm searching for a tutorial
yup
iirc https://docs.applio.org has a tutorial
why does it even say "Don't put it in a folder with privileged access.
Don't run the run-install.bat as an administrator."
this is scary xD
huh lol
?