[Tsundoku]

Fuck, Maxi Jazz died.

https://www.news.com...bf509d288055d80

This post has been edited by Tsundoku: 24 December 2022 - 06:38 PM

[Azath Vitr (D'ivers]

Free Bell Swarm virtual instrument for free Spitfire Labs player:





https://labs.spitfir....com/bell-swarm

Free Christmas bells for free Kontakt Player:




https://sonixinema.c...christmas-bells


You can get a free instrument for the free Sine Player with the OT 25 euro voucher: hh-ot-25-2022

It will only work on a single item though. Many good free options (anything 25 euro or less; VAT is applied after the voucher, but any % of 0 is 0...). (Scroll down and click on 'Instruments' to see them.) Here are some examples:

Modus - male cinematic choir, low flute ensemble, Klingon choir
https://www.orchestr...modus-jeffrusso

Salu (recorded in Estonia) - Kannel (it's like a Kantele, zither), Evolving (experimental drones), kannel harp and piano ensemble
https://www.orchestr...ollections/salu

Phoenix Orchestra - Pipa, gaoyin suona (brass), xun in D (ancient ocarina), zhongruan (moon guitar)
https://www.orchestr...oenix-orchestra

Andea - Ronroco (mandolin), charango (lute), cuatro (guitar), pinkullo 1 (flute)
https://www.orchestr...llections/andea

LA Studio Sessions - vintage electric bass, banjo, tambourines
https://www.orchestr...ons/la-sessions

Metropolis Ark 1 (The Monumental Orchestra) - Kornmesser Guitar 2, Kornmesser Guitar 1, Troppaur E-Bass
https://www.orchestr...etropolis-ark-1

Metropolis Ark 2 (Sounds of the Deep) - Hochbaum Organ, Thekla Harmonium, Roon Piano
https://www.orchestr...etropolis-ark-2

Metropolis Ark 4 (Elite Orchestral Forces) - Marimba xylophone and piano ensemble, tuned tomes and timpani, percussion ensemble
https://www.orchestr...etropolis-ark-4

Metropolis Ark 5 - Flugelhorn ensemble a4, low brass mellow, brass ensemble
https://www.orchestr...etropolis-ark-5

Berlin Percussion - timpani baroque mallets, piatti (that's Italian for cymbals), aux percussion
https://www.orchestr...rlin-percussion

This post has been edited by Azath Vitr (D'ivers: 25 December 2022 - 12:47 PM

[Azath Vitr (D'ivers]

'Featuring:

Solaria
Cong Zheng
Kevin
Xuan Yu
Asterian'

(of course these are all AI vocalists---naturally... though the chatter at the beginning is all too human)

[Azath Vitr (D'ivers]

New free 'Emotive Choir' for free Kontakt player:

[Azath Vitr (D'ivers]

'This Is the Band That’s Supposedly Saving Rock and Roll?

[...] Måneskin’s cover of the Four Seasons’ “Beggin’” became inescapable on U.S. radio—and, as notably, was globally TikTok’s second-most-played song of 2021. Some of the group’s biggest influences—Iggy Pop, Tom Morello—have become collaborators and fans, and next month, Måneskin may well win the Grammy for Best New Artist. The rise of a capital-r Rock Band in an era driven by hip-hop and electronic pop raises a question: Is Måneskin’s popularity a fluke, or is it a sign of some deeper shift in mainstream tastes?

[...] On song after song, [...] lay down a samey punk-disco thump, over which Raggi’s in-your-face scribblings and David’s blustering vocals sit like a funny hat. [...]

[...] the album’s redundancy has the odd effect of calling into question the guitars are back! narrative that the band seemingly invites. Certainly, some rock aesthetics have shown up on the pop charts lately, including new takes on indie rock (Steve Lacy), emo (Olivia Rodrigo), and Mick Jagger cosplay (Harry Styles).

But Måneskin’s songs are so plainly recycled, so brazenly mediocre, that the notion of the group igniting some culture war between rock and pop—and with it, stereotypes about realness and phoniness, passion and product—seems tragic at best. [...]

So why is Måneskin breaking out? “They are a television phenomenon,” [...] Rock can bubble up in the same way that Kate Bush or Disney show tunes can: through memorable televisual exposure.'

Måneskin Looks a Lot Cooler Than It Sounds - The Atlantic







Hmm, IDK... 'Tom Morello is a fan' of something 'so brazenly mediocre'?

[Tsundoku]

David Crosby died.

[worry]

Gonna miss that old goat.

[Maark Abbott]

"AI voices"

Just call it what it is, ya sausage.

It's Hatsune Miku. It's vocaloid. Welcome to the weeb dimension.

[Azath Vitr (D'ivers]

Maark Abbott, on 20 January 2023 - 08:36 AM, said:

"AI voices"

Just call it what it is, ya sausage.

It's Hatsune Miku. It's vocaloid. Welcome to the weeb dimension.

IDK if you've bothered listening to the examples, but the quality level of the AI generated vocals is exponentially better than Vocaloid.

Vocaloid is sample-based, and consequently takes up a large amount of disk space; Synth V is based on neural networks, and takes up very little disk space.

The AI vocals sound so much more natural and more intelligible than Vocaloid (English language Vocaloids still require subtitles to be understood) that they're about as similar as a medieval windmill and a modern nuclear power plant.

Synth V sounds good enough to 'fool' professional musicians into thinking the singer is human. Vocaloid... not so much.

Synth V:




Vocaloid:





You're reminding me of what Moby said about randomization of drum hits for 'humanization'---he claimed human drummers wouldn't be needed any more. Sounds just like them. Even though it was completely random. Neural network based humanization trained on actual drummers is a much better approach IMO. But the difference is probably less obvious to most people.... But fear not, people will still want to see and listen to people hitting stuff. When I was watching Whiplash, in the scene where the sadistic drum instructor is trying to get him to drum as fast as possible, his hands are bleeding etc., the obvious thought was---why not just have a machine do that (which it can do easily)? Why try to do what machines to so much better? And the answer is partly in the messy spectacle of humans trying to do such things. Though now with AI machinic productions can be imbued with a convincing simulation of the same emotions---which may eventually be optimized to be as convincing as possible, more convincing than human productions. But still (at least some) people will want actual humans to do it (even though crude drum machines took over most of popular music a long time ago). And some of them will want to watch....

[Maark Abbott]

The only time you can't tell if computerised drums aren't an actual drummer is is a drummer has programmed it. Otherwise, as with all things computerised / AI, it's not hard to spot at all. Convincing fakes will always remain fakes.

[Azath Vitr (D'ivers]

Maark Abbott, on 20 January 2023 - 01:46 PM, said:

The only time you can't tell if computerised drums aren't an actual drummer is is a drummer has programmed it. Otherwise, as with all things computerised / AI, it's not hard to spot at all. Convincing fakes will always remain fakes.

So when you can't tell them apart in a blind test will you finally bend the knee to your robo-overlords and embrace the freedom that the Basilisk brings? It's only droit. (Honi soit qui mal y pense....)

(In case it's not clear, I'm kidding about the dark superhuman Lord I serve Basilisk....)

[Gorefest]

I am sure there will be a market for it in certsin sections of music, e.g. pop music. Let's be fair, it is not as if people like Britney Spears and Beyonce write their own music or lyrics. There will undoubtedly be a market for cheap AI generated radio filler and it will mostly hit the section of artists who currently rely on autotune and shows like X factor. It shouldnt hugely affect touring artists and bands with a loyal following. If anything, you have to hope that current and new artists will find innovative and creative ways to use this new tech. Like people did with drum machines, electric guitars, amplifiers and grammophones. At worst it puts some mediocre singers out of business, at best it opens new creative avenues for artists to utilise.

This post has been edited by Gorefest: 20 January 2023 - 05:12 PM

[ContrarianMalazanReader]

There's this one song from the Black Crowes called "Live Too Fast/Mercy Sweert Moan", a snippet of which is heard as the final track of 1998 reissue of their first album Shake your Money Maker, but the band has never released an official studio recording of the full song, and I've read elsewhere that they've only performed the song live twice. In fact there's a YouTube video of them performing the song live which is the only way I know of you can listen to the full song. Not even the 30th Anniversary edition of SYMM has the full song.

I wonder why the Robinsons have never released the full song in any official capacity.

This post has been edited by ContrarianMalazanReader: 20 January 2023 - 06:09 PM

[worry]

Hmm, I'd be shocked to find that Beyonce wasn't the driving force behind most of her post-DC output.
In general, I'd imagine AI-generated tunes will remain a novelty even if eventually a few break through. I do agree it may become part of some producers' toolkits, but that it's not going to be an outsized existential threat to anybody.

[Maark Abbott]

That's all they are and all they will be - a novelty. A computer program mashing two things that someone fed it together cannot be creative, will never have the capacity to be creative. You might get some amusing noise, but it's the Shakespeare's Monkeys at Typewriters scenario.

[Azath Vitr (D'ivers]

Maark Abbott, on 21 January 2023 - 08:20 AM, said:

That's all they are and all they will be - a novelty. A computer program mashing two things that someone fed it together cannot be creative, will never have the capacity to be creative. You might get some amusing noise, but it's the Shakespeare's Monkeys at Typewriters scenario.

Again, that's not how generative neural networks work: they're not 'mashing two things that someone fed it'. They're learning the underlying patterns to generate novel examples from (generally) noise. Typically they effectively infer what constituent properties are significant (for example, in visual art brightness or weirdness, as well as many less human explicable properties or patterns), treat each property as a dimension, and construct an n-dimensional latent space (for Stable Diffusion it's > 500 dimensional).

But you did identify two obvious ways it can be creative: through randomization within the bounds of a style or genre, and by combining different styles or genres.

If trained on the creative evolution of a specific style, it can continue that creative evolution. It can also learn from a wide variety of genres to construct a general space of human 'musical possibility'. So instead of being uniformly random, it can randomize in a way that fits the inferred trajectory through its space of musical possibility.

Computers are so fast now that the problem isn't so much randomly generating the complete works of Shakespeare---in a world where Shakespeare never existed---as it is recognizing that it has. One option is to put vast quantities out into the world and have people rate them, explicitly or implicitly, and then train on that feedback.

More generally, neural networks can approximate any optimization function based on the training data. So they can be optimized for any measure of 'creativity' as applied to patterns inferred from the training data. (For example, neural network based chess computers excel at creativity, coming up with effective moves no human would have thought of. They do this partly by playing against each other a large number of times, with the optimization essentially being 'win'.)

However, they may be limited by what is included in the training data. Ironically this might exclude new technologies or aural aspects of human experience that have yet to be incorporated. Though given wide enough variety---including exposure to a wide range of aural phenomena that humans have liked in the past---it could infer much of the space of possible sounds that humans might like, or might want to incorporate in music (as so many already have with the integration of samples, noise art, chanted, declaimed, or spoken parts, etc.).

But actually simulating even a single human brain with anything approaching full accuracy is beyond the scope of current computers. So is simulating the whole of human culture and experience as it changes through time. However, how much actually 'needs' to be simulated to effectively predict what humans will like is an open question. Again though, AI can dynamically adapt to changes in taste by receiving feedback from humans on its ongoing releases---what people listen to (and continue listening to) or rate highly, etc.

[Serenity]

tricot - #Achoi

[QuickTidal]

Fuck...this song is in my bones right now....

[QuickTidal]

I'm here to take you back to the 80's for a minute (this song has been on every driving mix playlist I've ever made)

[Azath Vitr (D'ivers]

New heavy metal AI voicebank based on Ninezero (... with his consent and extensive participation):




'Australian rock/metal singer, songwriter and musician[... living in Japan ...] Ninezero joined forces with some of Japan's most famous musicians to form Maziora The Band. The line-up is Ninezero-vocals, Kentaro-guitar (gargoyle), Yoshito Onda-bass (Zamza, formerly from Judy and Mary, Presence) and Himawari-drums (Dustar-3, formerly from Sex Machineguns).'

Ninezero - Wikipedia