AI (improbable) songs
Page 1 of 1
HubU
VIP Member



Posts: 11256

PostPosted: Sun, 20th Aug 2023 00:38    Post subject: AI (improbable) songs
Quoting vurt & Amadeus from the void:

Amadeus wrote:
vurt wrote:
AI-Phil Collins sings Beat it



....one more Very Happy





Born too early to explore space but at least born just in time to see the golden age of AI - could be worse.



"Music washes away from the soul the dust of everyday life." ~Berthold Auerbach
Back to top
Amadeus




Posts: 2329
Location: Yes
PostPosted: Sun, 20th Aug 2023 00:59    Post subject:
Back to top
monk3ybusin3ss




Posts: 11155

PostPosted: Sun, 20th Aug 2023 01:48    Post subject:
Back to top
Shocktrooper




Posts: 4499

PostPosted: Sun, 20th Aug 2023 02:16    Post subject:












Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Sun, 20th Aug 2023 02:39    Post subject:
monk3ybusin3ss wrote:
https://www.youtube.com/shorts/cIQ4lf6vkJo


that's not AI, it's a really good voice actor.

The cartman AI above is great, best with him so far Smile
Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Wed, 6th Sep 2023 11:08    Post subject:
my first tests, pretty fun Laughing

Back to top
jackbomb




Posts: 2511
Location: Tortilla de patatas
PostPosted: Wed, 6th Sep 2023 15:32    Post subject:
Man, that suburbia-wario was Laughing

Are those your trained models? If so, good work!


B650 Tomahawk / 7950x CU -20 5.5GHz / PS120SE / 2x16GB DDR5 6000 CL30 / 3080 OC
Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Wed, 6th Sep 2023 15:49    Post subject:
No, i haven't trained any myself yet.

Another one

Back to top
TheZor
VIP Member



Posts: 5991

PostPosted: Wed, 6th Sep 2023 17:37    Post subject:
vurt wrote:
No, i haven't trained any myself yet.

Another one



Damn, that was eerily cool ! Smile Good fit ! A lot of it depends upon pairing a good singer with the track, but when you're using a good artist, the end result is usually okay at the very least Very Happy

I'm conflicted about AI and arts, but I'll take those AI covers any day of the century vs those terrible ass covers that you'll hear on whatever talent show Razz


R5 5600X - 3070FE - 16GB DDR4 3600 - Asus B550 TUF Gaming Plus - BeQuiet Straight Power 11 750W - Pure Base 500DX
Back to top
Dromaium
Banned



Posts: 462
Location: I don't identify as Banned, Ronhrin, Citizen.
PostPosted: Wed, 6th Sep 2023 17:40    Post subject:
vurt wrote:
No, i haven't trained any myself yet.

Another one



What do you use to train your musical models?

I have a PB musical collection, now I'm curious what I could do with it.
Back to top
DXWarlock
VIP Member



Posts: 11422
Location: Florida, USA
PostPosted: Wed, 6th Sep 2023 18:46    Post subject:
Dromaium wrote:

What do you use to train your musical models?


Same question, what you using vurt? Would love to play with it.
I tried this RVC:
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI

But didnt get good results.


-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf

Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Wed, 6th Sep 2023 19:21    Post subject:
I use RVC v2. I was surprised how fast it is, for the black celebration lyrics it's perhaps 8-10 seconds to process it with new vocals, with the best quality Surprised
Back to top
DXWarlock
VIP Member



Posts: 11422
Location: Florida, USA
PostPosted: Wed, 6th Sep 2023 19:27    Post subject:
Was you using a collab or local windows GUI?
All I can find for v2 is collab options.


-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf

Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Wed, 6th Sep 2023 19:39    Post subject:
Back to top
Dromaium
Banned



Posts: 462
Location: I don't identify as Banned, Ronhrin, Citizen.
PostPosted: Wed, 6th Sep 2023 19:41    Post subject:
vurt wrote:
I use RVC v2. I was surprised how fast it is, for the black celebration lyrics it's perhaps 8-10 seconds to process it with new vocals, with the best quality Surprised


How long does it take to train?
Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Wed, 6th Sep 2023 19:42    Post subject:
again, i don't train at all. i just process with already trained voices that i find. processing is perhaps 8-10 seconds on my 3090.

i use spectral layers 10 to remove vocals (this is the slowest part, maybe 2 minutes), layer back in soundforge or cubase, but any audio program works of course where you can mix.

https://discord.com/invite/aihub

this is where you find 100's of models. it's always full. you have to try for hours to get in, at least i had to Sad so yeah, download all you can if you get in Laughing there's SO much shit, it's 60% anime characters, 30% rap. Just terrible. sad to see what people listen to these days, these people are likely mostly teens, i hope.

Got to read up on how to train my own model, that's when you can begin to get really great results, i doubt these people know what the hell they're doing for the most part, it's probably sloppy as hell.
Back to top
DXWarlock
VIP Member



Posts: 11422
Location: Florida, USA
PostPosted: Thu, 7th Sep 2023 16:17    Post subject:
vurt wrote:
I use RVC v2. I was surprised how fast it is, for the black celebration lyrics it's perhaps 8-10 seconds to process it with new vocals, with the best quality Surprised

It is pretty fast compared to RVC1, have to play with the setting to get good results.

And it did take FOREVER To slip into AIhub discord. To get there and realize I am old, I dont know who 90% of the voice models are made from or know what random weeb thing it is...haha

What I made so far:
Winnie the Pooh - Hurt.
http://sndup.net/wxjg



-

Spongebob - Gin and Juice:
https://sndup.net/g8gb/


-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf

Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
Back to top
vurt




Posts: 13586
Location: Sweden
PostPosted: Fri, 8th Sep 2023 14:33    Post subject:
Nice. You can tell its an "overlay." i've noticed i must up the feature retrieval rate to max, otherwise you can hear its just the original singer but overlayed with another voice, just slightly different tonally.

making my own models now. my first was a drum model, using it over voices is pretty cool, every word is a drum lol.. still tweaking it. https://sndup.net/s9fq/
(its Tintin, in Swedish)

I totally expect we will get something like a general music model which can just change a song entirely with different instruments. let's say you have a song done with acoustic instruments and you can convert it into a techno banger (if you have a techno model) or use a model trained with movies scores and you will have something else entirely. This will get sci-fi level crazy Smile
Back to top
Neon
VIP Member



Posts: 18918
Location: Poland
PostPosted: Fri, 8th Sep 2023 14:57    Post subject:


This one is still the best I heard. Absolutely incredible.

That last chorus is beyond amazing.



This is also really good:

Back to top
Frant
King's Bounty



Posts: 24555
Location: Your Mom
PostPosted: Sun, 10th Sep 2023 01:09    Post subject:
Neon wrote:
https://www.youtube.com/watch?v=3_GpGbwxuoc

This one is still the best I heard. Absolutely incredible.

That last chorus is beyond amazing.


My fave part is when he sings "the hot dogs go on". Very Happy


Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn!

"The sky was the color of a TV tuned to a dead station" - Neuromancer
Back to top
Neon
VIP Member



Posts: 18918
Location: Poland
PostPosted: Sun, 10th Sep 2023 08:05    Post subject:
Back to top
ixigia
[Moderator] Consigliere



Posts: 65002
Location: Italy
PostPosted: Fri, 29th Dec 2023 17:59    Post subject:
Covers are impressive, but AI-made original songs aren't too shabby either:


(AI song/human singer hybrid)


(full AI)

Pretty damn spot on Very Happy
Back to top
madmax17




Posts: 18477
Location: Croatia
PostPosted: Sat, 30th Dec 2023 00:38    Post subject:
Back to top
madmax17




Posts: 18477
Location: Croatia
PostPosted: Mon, 28th Oct 2024 22:26    Post subject:
Back to top
Page 1 of 1 All times are GMT + 1 Hour
NFOHump.com Forum Index - Music Waves
Signature/Avatar nuking: none (can be changed in your profile)  


Display posts from previous:   

Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB 2.0.8 © 2001, 2002 phpBB Group