AI Generated Music is INSANELY GOOD! - Google's MusicLM

Google Research introduces MusicLM, a model that can generate high-fidelity music from text descriptions. See how MusicLM casts the process of conditional music generation as a hierarchical sequence-to-sequence modeling task, and how it outperforms previous systems in audio quality and text description adherence. Learn more about MusicCaps, a dataset composed of 5.5k music-text pairs, and see how MusicLM can be conditioned on both text and a melody. Check out this video to see the power of MusicLM: Generating Music From Text! #GoogleResearch #MusicLM #MusicGeneration
▼ Link(s) From Today’s Video:
✩ MusicLM: google-researc...
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ AI LINKS MASTER LIST: www.futurepedi...
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.c...
✩ Second Channel: / @matt_pie
-------------------------------------------------
Thanks for watching Matt Video Productions! I make all sorts of videos here on KZbin! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!
All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.
-------------------------------------------------
► Business Contact: [e-mail]

[@TheTunaBoi]

All songs: Audio Generation From Rich Captions: Song 1: 3:29 Song 2: 5:13 Song 3: 6:36 Song 4: 7:44 Song 5: 8:59 Song 6: 9:37 Song 7: 10:37 Song 8: 12:00 Song 9: 12:47 Song 10: 13:27 Long Generation: 13:57 Story Mode: Time to meditate: 14:34 Time to wake up: 14:45 Time to run: 14:56 Time to give 100%: 15:11 Electronic song played in a videogame: 15:22 Meditation song played next to a river: 15:30 Fire: 15:45 Fireworks: 15:57 3rd one: 16:10 Text and Melody Conditioning: 16:34 Painting Caption Conditioning: Painting 1: 17:00 Painting 2: 17:19 Painting 3: 17:36 Painting 4: 17:52 Painting 5: 18:12 10s Audio Generation From Text: Cello: 18:33 Hip hop: 18:44 Country: 18:55 Escaping prison: 19:10 Club in the 80s: 19:19 Club in the 70s: 19:22 Club in the 60s: 19:24 Club in the 50s: 19:27 Futuristic club: 19:30 Club in the 00s: 19:36 Accordion death metal: 19:48 Accordion techno: 19:59 Generation diversity: 20:16 THANK YOU SO MUCH FOR ALL THE LIKES OMG THATS THE MOST LIKES I EVER GOT

[@ki-zenn]

Omg ty for this

[@TheTunaBoi]

@@ki-zenn np

[@shxdyneo7298]

Thank you for letting me watch what I came here for without his dumb fucking opinions

[@GEMSofGOD_com]

My most been over 8000 for "selling" fake Hublots 😂

[@NickNov]

You deserve 10k likes for this

[@ofulgor]

Just like AI images, AI musics are evolving so fast. This is insane.

[@sasufreqchann]

you didnt listen to others ? like other ai ?

[@eh_eye]

AI everything is evolving at uberspeed

[@Anti-HyperLink]

And musicians are gonna start whining.

[@TheMusicolophile]

Evolving. LOL.

[@TheMusicolophile]

@@Anti-HyperLink Definitely don’t want those fleshy meat puppets whining, do we?

[@nickbarton3191]

The existential danger from AI is that we'll be immersed in a sea of derived banality and we'll die of boredom.

[@luckyankraj]

Well as a musician, this is impressive and scary at the same time.

[@allwheelmodeltoys491]

This A.I. music is nothing but a copycat

[@jadenhancock]

Same

[@zeenagual1962]

dont worry , it sounds cheap and cheesy , like the ppl who believe in it !!!

[@MorbiusBlueBalls]

i think musicians are in the safe space in ai advancement. a person can become the next hans Zimmer, ai can't. ai can't even tell if the music it created sounds good or not

[@illcaponesok833]

@@MorbiusBlueBalls we said the same thing about AI images. The AI doesn't need to know if it's good, we do. We say what's good or not and it listens.

[@joebender9052]

Language model ethics concerns: "That's too convincing. It could be used for mass propaganda." Music model ethics concerns: "That bass line is too funky."

[@joebender9052]

On a serious note, the main issue at the moment is the audio quality. If there were another model that could scale up the resolution, I would listen to these songs along with my other favorite music.

[@javier_afonso]

I usually don't participate on youtube conversations but I'm so horrified by the points of view from a lot of people in every single ai art generation that I need to say it out loud (or type it). The fact that you (and countless others) don't see any ethic concern is a really worrying sign. Let me spell some concerns for you. First of all, the AI are fed with works made by real people, not machines, who devoted their lives to perfect their art. They haven't been asked for their permission, not just that but after having conversations with many professionals and talented amateurs I've noticed that 99% are angry and borderline depressed. Not only someone's using their work, which took them a lifetime to develop, to feed a machine meant to replace them. Not just that, but all of a sudden their live's purpose is swept from underneath their feet. I don't know if you know many artists, but it's not just a profession, for most it's a 24 hour thing, an obsession and a goal in life. Then there's the fact that you think about music as something without any effect on society in general. Just think about what would've happened to the 60's without the counterculture movement, or the effect 1984 has had on people's perception of government control. Art is not just pretty meaningless things to stare at, it can be the glue to unite a generation, it could be a mirror to represent problems so abstract that are difficult to articulate. In some cases it's the physical manifestation or an amplifier of a collective's emotions,, principles, lifestyle... and also what will be left behind once that collective dissapears. Letting corporations manage that power (which has been happening way before AI took shape, but is now reaching a point of no return) implies a significant mean of mass propaganda, because in some ways it's a mean to control a group's identity and way of thinking. DOn't forget burning books and killing artists is one of the first things authoritarian regimes do. Nowadays it sounds naive to say that, but that's because art has been steadily deployed from it's meaning and it's mostly shallow and hollow. But AI will be the nail on the coffin. Then there's the economical aspect, who's goign to benefit from a model trained on creative workers? Mainly corporations and people who didn't create anything. Then the aspect of denying future generation of their artists. Who's going to go through all the hardships that are needed to become a really good artists if a machine can give you something with just a loose description? What's going to be the motivation if the final result is irrelevance and being forgotten in the noise of millions of meaningless and derivative AI generated images/songs?. Then there's the worrying homogenization of talent. Fifty years ago there were a lot of jobs that required different ways of thinking. Although I'm not that old I've seen the change from that model where an artist,a writer, an administrative... they all had to have their own set of skills and a unique way of thinking. Now the only profile needed for every job is a programmer. You can record and album without knowing a single thing about music but you will have a lot of problems if you don't know a single thing about computers. Because everything tends to be optimized for a certain type of person, not the target professionals. And although I have many friends who are wonderful human beings and work in the field, I've seen with the years that as a collective is quite a toxic one, assuming that everyone should have the same thought process as them and transforming any profession into theirs, even when that means a negative impact on the target professionals instead of helping them. Then, there's the question. WHY. There's countless things AI could be doing to benefit our world (alphafold comes to mind) and in my opinion progress means making our lifes better. I don't see how this makes anyone life's better. Certainly not for musicians, not for listeners who may be fooled into it but in the end will have a shallow experience instead of a real human connection (which music is all about). Meanwhile, the jobs that are left for us are becoming incresingly more depressing, with positions that are based on repetitive tasks that an AI should be doing, working hard in something unfulfilling and meaningless jobs. I think it's really cruel to live in a world where humans have to do things they hate for a living while computers do the things we wish we could be doing.

[@joebender9052]

@@javier_afonso I'm actually also a musician. Nothing that I've made has been listened to by very many people, and I've never had any hopes of making money from my music. I still will continue to create music because it is part of what gives my life meaning, and the presence of other music in the world (whether created by humans or AI) can never prevent me from making my own music. Music models like this one can actually help me by giving me ideas for chords, melodies, rhythms, and even timbres to use. If you only consider music meaningful when it is famous and appreciated by millions of people, then that is out of reach for virtually all musicians already. As for concerns about people supporting themselves through their music, and how AI will affect that, this video addresses it well: kzbin.info/www/bejne/j5TZiYGaoLCErqc My original comment was a reference to when OpenAI wouldn't make their original GPT-3 public due to safety concerns, and how music doesn't have that same specific type of danger. I was only thinking of instrumental music, though, and you have a good point that vocals can carry propaganda messages. I do recognize the ethical concerns of training models on data without permission, and also the potential threat to many jobs.

[@Patrick-ryan-collins]

Music is far more influential than text

[@herbie_the_hillbillie_goat]

@@javier_afonso There's no need to worry about ethics here. If you don't want people to use your work, simply don't make it available to the public. It's not uncommon for musicians to draw inspiration from each other and create something entirely new. Your concern seems to stem from a fear of the unknown, but as a composer myself, I don't find the use of AI threatening at all. I can still create music regardless of whether others can or not. Additionally, it's irrelevant whether a piece of music was created by a human or a computer if someone enjoys it. The fact is that the future is here, so you can either dwell in despair or embrace it. As for me, I'm going to focus on making music.

[@limelorax]

an arpeggio is separating a chord and playing the notes individually. The buildup is an increase in volume and is called a crescendo.

[@therantingboy]

And it's opening a low pass filter on the synths

[@kimrunic5874]

@@therantingboy no, a crescendo is necessarily an increase in volume. A lo pass filter affects frequency, not amplitude.

[@Spuzzmacher]

Honestly. My analog syth from 1978 could do this. I guess it's easy to be impressed when you don't know anything lol

[@TheArthurAbbott]

Well said. He probably was thinking of a crescendo.

[@avi7278]

I never knew that I actually didn't need real words in my music. Now I know what it's like when non-english speaking people listen to music in English. It actually doesn't really matter.

[@games528]

Lol have you never listened to non-english music before?

[@avi7278]

@@games528 LOL, Not to brag or anything but Im native bilingual in French/English and fluent in Spanish. But also these faux lyrics have a very English form and sound to them which is why I thought of that example.

[@PuppiesAreNice.]

50% mind blowing 50% nightmare fuel

[@cactusp00p]

Google probably needs to start offering public use of their AI developments.... Microsoft really helped push the competition for Google to start releasing demos out for the public.

[@NicVandEmZ]

Agree

[@thedoctor5478]

I won't be holding my breath

[@sasufreqchann]

their trash ai is nothing worth attention fr fr .

[@NicVandEmZ]

@@sasufreqchann also most of there projects just get axed in end and end up in google graveyard

[@tentrilliconlp7748]

They probably never will.

[@fizzypizzel6477]

What I'm most excited for when it comes to Ai ranked. 1 Text to Virtual Reality (sandbox) 2 Text to Film (with sound and visuals) 3 Text to Music (especially combining existing music or making variations of it) 4 Perfect Ai Photoshop 5 Improved Text to Image

[@holdthetruthhostage]

Virtual Reality is dead man, because social media exists, why would you try to build your audience as a 3D model anyone can do it's far less useful than AR man

[@cihiris2206]

I'm hoping for Text to 3D modeling and also Text to VR too. VR is not dead.

[@gecko5873]

@@holdthetruthhostage vr would still be good for games

[@Kudjr222]

@@holdthetruthhostage the quest 2 is kinda making vr relevant again.

[@hombacom]

Two kind of persons, one that thinks that AI should do everything for you and one that see it just as a tool among many - which both can disturb you to finish anything

[@arnelilleseter4755]

When I was a kid, influenced by british and american music and TV, we would often pretend to speak English even if we didn't understand a word. The vocals in this sounds excactly like that.

[@matthiasheymann]

For those listening to all the commercialized trash that is produced these days, this certainly sounds indistinguishable.

[@finnfransen_official279]

Thanks dude!

[@markkennedy9767]

What the painting sounds like to the AI is another level of intriguing. Basically digital synaesthesia. Just the whole idea of extending meaning is actually quite interesting. Truly novel.

[@Rctdcttecededtef]

It would be cool if one day it were possible for AI to generate music based on music you already like

[@tonyanimations5377]

Its simple, you just describe the music you like and it will generate a similar music.

[@virgilxavier1]

6-12 months?

[@Rctdcttecededtef]

@@tonyanimations5377 you're right but I feel as though having access to the entirety of a playlist would be working with more data than what you might be able to provide in a text prompt but I could be wrong. I don't know how this ai functions

[@VperVendetta1992]

It kind of can already from what is shown. They have music to music, just like we have img to img in Dall-e and stable diffusion. You load a music and then ask it to create a variation.

[@Rctdcttecededtef]

@@VperVendetta1992 I haven't seen much in the ai music generation space other than what's presented here and jukebox ai

[@kennyrosenyc]

1970's - 'In the future computer will do all the work and humans will focus on art'. Reality - 'Computers make all the art and humans do all the work'. It's a good thing these 'AI programmers' are scraping the internet for material to use in teaching their AI systems to create this stuff thereby allowing the original artists to sue them for copyright infringement. You can't take somebody's work without asking first.

[@Michaelkaydee]

Had gotten bored of tech until AI got to where it is now... I can't get enough... spending hours on end researching it and where it's going and messing about with it... love it.

[@doingtime20]

This is insane, AI related things are going so fast it's hard to even keep up.

[@michaelpettersson4919]

Welcome to the craftsmans experience of industrialisation. There was no pity shown for the craftsmen then. Automation have made masses unemployed, no pity shown to them. Welcome to the working class experience for more then a century. Tell me, why should we show any pity now of a sudden? Time to learn to do...Nothing. Everything will be done better by machines, even art.

[@fanban2926]

@@michaelpettersson4919Just because pity was not shown before does not make it right.

[@andersonsystem2]

Wow, this is freaking amazing. Google should release it as a beta to the public. Google has amazing AI, so they need to get themselves in the game. The music is actually great.

[@kuromiLayfe]

think majority of the ai systems they are showcasing currently can only run on super computers on quantum processors and with Petabytes of storage and almost a Terabyte of Ram/VRam

[@VperVendetta1992]

@@kuromiLayfe You can run stable diffusion with great results with modern Nvidia video cards, so no. This is not that different in terms of resources. And they can anyway offer an API endpoint like OpenAI.

[@rizizum]

@@kuromiLayfe You'd only need that if you want to train these models

[@MentalPistol]

as much id like to use this, dont think we'd be able to try it anytime soon, if ever,

[@kuromiLayfe]

@@VperVendetta1992 but can modern day home systems generate 15 second video at 12 fps 720p in 4 seconds per video? or would you need to get your hand on A100 GPU’s to get even close

[@johnnywoodstock]

Wow man I'm dumbfounded that is insane thank you for making this video and informing us about these advancements in AI technilogy

[@Vansafe0]

The music quality seems very similar to how the quality of Midjourney was to Ai art a year ago. It is neat, but not yet pleasant or mistakable for real music, but I think it may get there.

[@darkwraithcovenantindustries]

It's passable, but when you listen closely a lot of it kind of sounds like nonsense (i don't just mean the vocals). It's completely devoid of soul, and sounds like what an alien would make if you described music to it and it went off of that. It has an uncanny valley effect that when you listen to it closely, doesn't really sound like good music. It's interesting to hear non-musical people interpret it versus actual musicians who understand how actual music is created and composed.

[@Briman2052]

@@darkwraithcovenantindustries Agreed, it sounds crappy and soulless to a more "trained ear".

[@burkhardstackelberg1203]

The language usage in the songs reminds me a lot of the early diffusion models creating images who, when it came to text, rendered something resembling letters, then faulty texts - and now really get your text into the image. Give the models some years, and they easily will render a poem you give them into a song.

[@greglory]

Some years?

[@burkhardstackelberg1203]

@@greglory Even I wouldn't be surprised if it comes down to months - or a working prototype is running somewhere.

[@greglory]

@@burkhardstackelberg1203 no doubt. I've been trying to find an old article (either by Vice or Wired), in which the author went to an AI conference. The guys there told him they would trickle this stuff out, there reasoning was how it would effect employment. He thought they were being over enthusiastic, to which they replied it wasn't hyperbole...these tools were ready...that article was 2017 I believe.

[@dogisgone]

Months? Lol it's already been 1/3 of a year

[@WyteChinpira]

I used to dream of melodies and music when half asleep and thought it was amazing sounds of music... Maybe I can describe it in Ai and put it in real life now!

[@AltKaxREAL]

that sounds so wonderful!

[@tickledonions9483]

Just make it yourself. This is for lazy people.

[@WyteChinpira]

@@tickledonions9483 Exactly what I thought. I prefer raw instruments over clicking a button and boom.

[@jonathanrynjah]

Write those melodies down like most people do. Then play them once you get an instrument

[@jonathanrynjah]

Heck midi already exists.

[@bombasticbushkin4985]

Artificial for those who are looking for that. Definitely prefer real bands and natural sounds but these creations could saturate the sound waves and not necessarily for good. Not exactly rich orchestral sound or feel, power and quality of great band music. Lots of electronic pop synthetic sound.

[@cactusp00p]

The hip hop one is mind blowing. Sounded like an actually good kpop song.

[@wraser]

This would be very useful for content creators who need DMCA-free music for videos, podcasts and strems. So many people try to use copy-right free music in their projects only to get stung later on down the line when the published source music inevitably gets digitally watermarked or commercially distributed.

[@micheleduritto]

buy music licenses or use the free music that it is available on the internet (ever heard of creative commons?) no need for AI, there is plenty of materials....thank god for now that music AI is still pretty bad.

[@ChristianIce]

Technoaxe.

[@toddmaek5436]

The content creators will eventually be replaced as well

[@merlefi6162]

This is pretty interesting! I actually had lots of fun browsing the Music LM page for the classical painting inspired generated music. Some of the music for the paintings there actually sound creepy and weird (I like the one based on Scream tbh) but also very cool at the same time!

[@WeerdMunkee]

AI: The most creative thing ever invented. Also AI: An invention that practically ends creativity. Way to go humanity! Bang up job at this whole life/existence thing! 😃👍😒🙄

[@calebweintraub1]

Story mode is a promising idea…could imagine a corollary version of that for text-vid, LLMs, and down the road a bit-generative gaming.

[@Je-Lia]

The AI's singers sound like what I hear when I'm trying to understand wtf singers are saying! Lol

[@TheRandomRager]

Yeah sometimes its so hard to understand lyrics

[@PentagonDerZweite]

I can already see sandbox computer games using dynamically created background music, for ai made up story lines with fitting textures. What a time to be alive! Thanks for sharing, Matt!

[@FinalfixMusic]

Ridiculous. I can't believe this is how you imagine it will play out, think harder.

[@victormutunga2944]

It'll be a trip fr. Imagine a game where the only parameters u give the ai is the prompt for the storyline and how long u want to spend playing and it tailors the game to perfection

[@anarki777]

@@FinalfixMusic maybe you should think harder, bro. I can also see OP's vision playing out. Human's not required, better get used to it.

[@FinalfixMusic]

@@anarki777 Oh I have no doubt! IMO OP's imagination did not stretch far enough.

[@TheDefender123Plays]

Dynamic music is nothing new. It's not exclusively an AI related concept.

[@kewlnes987]

It is so incredibly bizarre and creepy but wondrous to be able to hear a computers musical thoughts. Had this same feeling with AI Images, especially with images of people who don't exist, but this hits different and harder. Music is so emotional, we all connect through it, hitting complex emotions unexpressable with words. Some of those melodies gave me that ponderous but chill mood about what is the meaning of life. I know the computer is mimicking our style through a vast amount of data, but, some of these samples made me empathize with a computer?

[@cesar4729]

Its improving. This have some vocals way pitchy, but is promising. Also 24K, instead 44K of standard quality audio.

[@LydianMelody]

Synthesizer V’s Solaria AI voice pack. 100% human sounding vocals. I used them in my last video but you can just look it up in KZbin if you want. Y’all are sleeping on it.

[@dockdrumming]

I am playing around with Synthesizer V now. I am going to create vocals for a track that I've made with it.

[@dickstryker]

I kind of like some of the weird artifacts. That flute sounded alien af and the rythm tracks were deliciously off-beat. I think there will be some very cool results from these elements being structured in a more musical and deliberate manner.

[@chiellazona5624]

All of them sound weird like you'll feel that it is not made by human, but there's this effect where you can actually enjoy listening.

[@theonlythingihavetosayis9333]

You act so reluctant to enjoy music.

[@rsstnnr76]

Pretty cool. Everything still has a twangy, midi sound to it. But I'm sure it will sound more natural before too long.

[@devin2489]

Is that a good thing?

[@lescakes]

It's amazing but kinda scary at the same time how the AI almost taking over most of what people are looking for

[@gartersnake3288]

5:34 Matt: is about to pause the song The song: Don't go away

[@ed.williams]

Note that, as the paper states, the music generation is limited to 24kHz. To put that into perspective, 16kHz is the telephony "narrowband" while 44.1kHz is standard CD sample rate. The MusicLM clips can't be used as-is for any serious purpose; you can't up-sample something that was recorded at a lower sample rate.

[@nathanlawrence2484]

Hey there. I'd like to thank you for making a video on this. I do have some criticism on the AI that you presented. I've been messing around with different AI thing since I was in Middle School, and I am presently graduated from college. One thing that I've learned about all AI is that even when they improved they are still minor Telltale things about their work or results that can be noticed easily. I've been following music for a little bit too, and one thing that I've noticed is that all musically I even this one seem to struggle with getting rid of issues with Tempo. Well this one is definitely much better than other AI that I've seen in the past, it's still has that trademark issue with keeping a constant tempo. I'm going to go out on a limb and theorize on how I think this AI works. Firstly more than likely they gave the AI a bunch of samples from instruments both electronic and not so that way it's able to use them appropriately with a lot of music a I typically don't have access to, and it was probably taught how a lot of different styles of music work etc. In every example that you show, the text prom is very specific, probably because they I cannot deal with off-topic or obscure or very large Concepts and ideas. More than likely through a lot of testing they've managed to figure out what the AI is best at and how to circumvent its quirks. Also, the vocals are trademark janky as heck. They aren't very good and probably won't be for several more generations.

[@TraderTimmy]

Above comment generated by AI

[@fanban2926]

@@TraderTimmyMaybe, but still correct for now.

[@futurepediaio]

gonna be super interesting to see the battle b/w google and msft/openai this year

[@NextGenart99]

A.I music is going to be crazy, imagine hearing a new hit song every day or shoot every hour. My favorite was the reggae I’m Jamaican and that was spot on and so good, especially the ending.

[@scottmasson3039]

This video, everybody in this comment section, the stunning success of chat gpt, everyone of my friends downloading AI art apps lately, etc proves to me one thing: we are done. Humans are over sooner than I thought. We think this is some cool new toy. We are just openly embracing this crap, even though we’ve been getting warned about the dark side of this stuff for decades. We love it already.

[@ryintel]

Sorry but all that “music” sucked. I have yet to see anything from AI that doesn’t have a stock feeling to it, stock copy, stock art, and now crappy stock music. Is it me or is this another Hyped tech “disruption”, Like Metaverse, VR, Crypto, the list goes on and on.

[@bw2937]

Agreed

[@cihiris2206]

I'm so excited for this! The possibilities are endless. I can't wait to use it to generate chill music for my VR social world creations.

[@nakedpear3558]

19:59 sounds like Markiplier’s old outro music

[@pheargoth]

As someone who has spent the better part of thirty years learning various DAWs and programming synthesizers (both hardware and software), This is some pretty clever stuff that the AI is doing. A few of those 30 second clips from early in the video inspired me to make full music, based completely on the short clip.

[@harryjones5260]

not that clever after all theres still only 12 notes

[@Pauluz_The_Web_Gnome]

OMG that is some creepy stuff: "Painting Caption Conditioning" -"Guernica - Pablo Picasso" An oh yeah, If you have a SUBWOOFER turn it up! The last 5 seconds, scared the hell out of me!

[@erics7004]

I'm a musician and I'm loving this idea. I play synthetizers, so, music is already going to this AI direction for a long time.

[@Flo-zo4mr]

This makes no sense. What has a synthesizer to do with AI?

[@bareknucklewarrior]

@@Flo-zo4mr init

[@Rollthered]

WOOOOO LETS GOOO, MY DAY HAS COME!!! I'm gonna be making so much samples for my music production with this. I make beats for fun and having AI to give me new ideas is like a huge blessing!!!

[@TheDefender123Plays]

It's certainly a convenience

[@davib8963]

Commercial art becomes a kind of "flower arranging" with this technology.

[@kato_dsrdr]

We'll have an AI capable of making whole songs from prompts complete with lyrics and voice synthesizer like a vocaloid.. IT WILL HAPPEN.

[@RockyFretz]

Almost...arpeggios are a series of notes like in a chord played in a particular pattern up and down...the build up is called a crescendo. :) im wondering if we added tempo indicators like 60bpm or 180 beats per minute what it'd do. Thanks for the great content. Im a semi retired musician and love AI stuff. Im writing 3 different KDP books in 3 different genres and 3 different age demographics and the bg music and providing the narration... all AI assited. Now, I wouldn't change much about how my life has gone these past 65 years for anything, but it's a whole new world, and Oh to be your age with these tools. I hope more young people are grasping the gravity of this tech and how it's going to drastically change their lives. Im even thinking that if i can make it another 15 years...there will be such movement forward in the fields of medicine, healing and longevity and gerontology...that I'll be able to take advantage of it and live long enough to see you at my age now! Man...that's gonna make inheritance stuff really interesting. :) Sorry to ramble, but I've been meaning to thank you for your hard work and great content. You keep it up, and you're likely gonna have an amazing life until you just get tired of it and "move on to the next thing" of your own free will! Now, That's longevity! ✌️

[@RealPapershaper]

I'm excited that we can finally tell the record labels that falsely claim the music for our videos to F off.

[@arxmechanica-robotics]

For music those are not very specific prompts at all. If you're a musician that doesn't tell you anything about a song.

[@devbites77]

I'm quite impressed with this as a start. Interesting times! For all those making quick judgements about the capabilities of AI keep this in mind. Don't look at where we are now (it's old research and initial attempts), but take note of the trends to imagine what may be possible in the near future. I think we'll all be surprised or shocked when we pass a threshold. It will seem like there is a sudden leap in capabilities. It won't take much.

[@markkennedy9767]

I agree. It's not even about what's technically possible now. That's a given and it's a given that that is gonna improve unimaginably. It's now just about paying attention to what areas are subject to these huge changes. And that could be anything. Weird times.

[@mrlightwriter]

I see that you watch 2 minute papers too ;)

[@paulm3969]

Great prediction, things will improve.

[@jj8703]

I am so glad theres finally some music AI generators, something that has been untouched for a while. this one is actally really good, although i have to say openAI's jukebox is better with lyrics, as you can actually type them in. but google's AI is a lot better at making the actual song, and a lot easier to describe the song. but as expected, they dont release anything. but this is still really good

[@maxpro751]

The goal of their AI projects, is to make money and releasing their AI with no fundamental plan on how to make money isn’t wise. This AI music will probably be sold to professionals, and will be used as a tool to music creators. Not everything is free in life.

[@jj8703]

@@maxpro751 oh yeah i forgot google is like the biggest company on earth

[@tickledonions9483]

When everything is randomly generated, nothing will have meaning.

[@drowzy2309]

@@tickledonions9483 if things can be randomly generated, then that fact alone should bother you more than the actual random generation.

[@DansuB4nsu03]

@@drowzy2309 A good example of this is ourselves. Our genetics are randomly generated and thus influence a lot of factors about our looks and behavior, among other things. Thus, even humans themselves aren't safe from the burden of random generation. Just some food for philosophical thought.

[@justinwhite2725]

0:32 'calculating really fast' is all AI models do too. It's all calculus and gradient descent.

[@lemapp]

Two things come to mind. I remember around 2000 visiting the Computer Museum in Boston. There was a display of a computer that generated several genres of music each evening. The result were barely okay. You could recognize that it got the genre correct. Additionally, there was a study done of pop hits to determine if there were any similarities. There were, but it was a multi-axis analysis, it couldn't explain what made something a hit. But it could analysis a song and make a prediction. I can see how the AI is an extension of what was on display in the museum. I'm curious, with today's more powerful computer, if that could incorporate that hit analysis to improve the output.

[@jasonlin9268]

Where is the computer museum located at?

[@justincanton]

Remember when people use to either be creative, talented or both. Everybody is flocking to AI so it can do the work for them now days. The start of the demise of humanity, creativity and individuality. I see the initial appeal and yes, you can get creative using something like this up to a point in terms of directing prompt’s, but did you really create anything you hear (No). Stop and think about AI in 20 years, will you be in the kind of world you want to live in ? Will you even be needed? If something can do everything for you would you even need a body or freedom. Is AI taking over the world in more ways than we currently think, you know that we are always kept many years behind the technology that really exists. Maybe some celebrities are AI? Hmm nuts ha, then why are they freezing on live tv, and things like Katy perry’s eye malfunctioning on stage... Something to think about. 🤯

[@aaronbeyer3635]

I wonder if you can put your own chord progression into it or add your own notes so it builds off the melody you created. I make beats off Ableton so I'm concerned about this making my skill obsolete 😭

[@elujinpk]

Meh. Learn an instrument you’ll be fine. These talentless people will only fall into circles jerks of self praise like ai art. They are doing nothing. And are talentless. Learn a skill. This will separate people who are worth something in the future.

[@ChristianIce]

How can your skill be obsolete if another tool will be available for you too? Do you think this tool will work better in your hands or in the hands of somebody who can't tell the difference between a guitar and a bass? :)

[@elujinpk]

​@@ChristianIce They need you to describe skill. LMAO. cause AI people have none

[@Octamed]

@@ChristianIce A hammer isn't a tool if it just makes the whole house.

[@ChristianIce]

@@Octamed You mean you would listen only to music with an AI voice saying random stuff? Even in 1000 years only crazy people would do that, it's completly pointless.

[@Vtlkvtlkvtlk]

Bad.I have checked the samples , the pitch is always floating, it is unpleasant most of the time

[@RE_Mastar]

Have you taken a look at AVIA(AI composition assistant)? Its been around for a while and generates music using VSTs and midi.

[@cody8860]

AIVA is overrated. Name a real musician or producer that has used it.

[@brandonmatson7618]

I am looking forward to this kind of Ai NOT as a replacement, but as a way to make high quality samples for my music. Like, let's say I have a 10 second breakbeat pattern or whatever or loop. I can then ask an Ai to make 20 minutes of audio with it and only it. Gets lots of cool variations. Then use it in my music. The music here sounds like someone who just started making music about a month ago. But of course there is no way to go but up. It's gonna be awesome... well until it can make better music than us in a few seconds.

[@Whats_newss]

It's inevitably going to replace and inflate the entire market though.

[@Chazzvc]

Text to 3D would be amazing. image to 3d would be amazing too. Where I can draw a character and then have ai generate dozens of 3d models that are similar to that 2d drawing. Also the ability to use 3d models to image where the Ai training can understand everything about the 3d model you feed it and produce perfect results that are consistent. The speed at which manga creation will reach in the next 5 years is going to be mind boggling. Mangaka in japan might actually get to sleep for once, Imagine that.

[@MagicBus101]

That's what digital artists are for! These AI programs are hurting and stealing from artists and gives no credit to the artists from where the art is taken! AI sucks in this sense

[@aidenlinz8422]

I saw something about AI making music. OpenAI did something with that but they kinda failed badly with it. But google seems to go hard.

[@brose2323]

It's like that episode of Rick and Morty where Jerry gets stuck in a simulation. " now playing, human music."

[@S0nyToprano]

I love this AI stuff. There was a print that no artist was willing to do a commission I offered to pay. Easily within market rate might I add. The reasons were many: “I don’t want to”, “that doesn’t inspire me”, “I’m too busy”. So I decided to try one of those AI generators. Told it what I wanted and after a couple tweaks to my wording, it spat out exactly what I wanted. It was absolutely perfect. No BS haggling and e-mails back and forth with artists. I’m sticking with AI moving forward. It doesn’t complain. It just does what you ask.

[@cn6519]

Yeah, but your art will also be the same style, and nothing inspirational. It is great for now because the tool is new, but come 10 years when everyone and their mama is using the same data set from 2020 or 2021, because most real artist jobs have been lost, eveything will still have the same inflections and noise.. Just like many commericals we see today, probably also heavily relying on computer imput or recommendations. Just remembert, the haggle is sometimes part of the process to become better. Everyone skipping to AI to try and avoid that, will make for a dull world in the end indeed.

[@Remour]

Society will get lazier with ai art, this person is proof of this.

[@guywhoisaguy67676]

Will it sign your guitar ? At the concert, Who's on stage ? Bunch of lip sync dancers and choreography pretending they wrote this sh*t ? And, you believing it ?

[@SkepticalCaveman]

I need more "accordion techno" in my life.

[@BlackRose369.]

You still have to come up with exceptional prompts also here

[@blakesun]

suddenly, all those musicians who made their living with background music in film, t.v. and commercials are going to find it hard to find placements.

[@Rctdcttecededtef]

Imagine being able to have it recreate songs and have them sung by different artists

[@Heheha329]

Not that hard just make it remove the vocals from the audio which we already javelin AI for that then use one of the AIs that can do vocals which we already have ,we need one that can search the internet and grab snipets of somebody singing to get samples for the voice copying AI (which exists already) and then just replace the vocals with that then a mixing AI

[@nestortomaselli6372]

Here is the thing, my entire life I've grown up around musicians. Even though I'm primarily a visual artist, music deeply inspires me and moves me to create my art. Any skilled musician I know, can compose better melodies than this in seconds by simply improvising with a synth or their given instrument. A DJ can remix and do all of the things this AI does already on the fly - and make it sound way, waaaaay better. Besides the novelty factor, most of these beats sound overly compressed, out of tempo, out of tune, the lyrics are unintelligible gibberish, and the sound lacks the soul & spontaneity that a human artist can bring into the mix. Most musicians I know would laugh at this even being called 'songs.' I think that (thankfully) human musicians will be WAY harder to replace by AI - not only due to the extreme copyright laws guarding the industry - but also the nature of the craft. Human musicians are already ridiculously prolific by the nature of how we interact with instruments. The process is not as lengthy and time consuming as painting or animating, for example. Sure, putting on a huge orchestral production is way more difficult and time consuming than jamming out a rock song, or even DJing at a club - but in any of these cases, the quality of the final product will always outmatch what AI has to offer because it has emotion. Emotion is at the core of music. In any case, this could replace people who create stock music online, but that kind of stuff is already pretty quick and inexpensive to make if you have a bunch of samples and a synth to put them together. These sounds are at that level, or even worse. At least stock music keeps a consistent tempo.

[@mitchcoppo]

I could see this replacing background music for ads, trailers, and eventually movie and game scores.

[@AlphaCentauri2]

At this point AI art and images are really good, AI video are okayish, AI conversations and information extract are good, in music it is not that good but still it is great. Currently AI is being companion to music, art, technology, law, engineers etc. but I think 3 years in future we will have very powerful AI that will be more then helper only. And I think that is true because is evolving really fast.

[@EgyptianEagle334]

Technology has evolved so quickly

[@wallyflint]

I think it is so wonderful that artificial intelligence can now write and record music. Just think, this means we humans don't have to bother with learning to play guitar anymore. But the real victory will be when it gains the capability of listening to it for us as well. Why just think of the time that will save!

[@grantezell9497]

Chat GPT, write me lyrics. Music LM, make a song with these Lyrics. Chat GPT, write a quick summary of this song. Perfect!

[@MagicBus101]

Lazy, pure lazy. If you don't have something to use your brain and abilities toward, why think or create at all? Sensless.

[@samgenia4245]

​@@MagicBus101 i feel like over the next few decades we humans will start to see an evolutionary branch off consisting of the humans that have relinquished their creativity and thinking abilities into technology, contrasting the humans who stayed true to discipline and faith in evolution and god. Unless the actions of the unconscious humans cause some sort of all around extinction, we will see a new breed of human that will be interested in nothing but Instant gratification, like sex and attention. Oh wait🤔...

[@Rsonixtunes]

Sarcasm is welcome.

[@CrimsonGrowlithe]

I feel, if you’re struggling to get an idea for music to produce and you’re a musician, this would be a nice starting point to expand upon because you’d be able to put your own human spin on it.

[@nomandad2000]

Na, soon it won’t need a human…

[@fanban2926]

@@nomandad2000Not soon, but still not all too far it could be much more automated

[@havemoney]

In the foreseeable future we won't have to expect anything more than the music we hear when waiting on the phone

[@x.4132]

Artists and creativity slowly dying. Super cool -yeah. People like you can't even imagine what this will lead, too.

[@wgxyz]

Seriously why though? Generative AI in creative fields is gonna destroy a lot of peoples meaning in life. Systems need to be in place to support these people

[@wgxyz]

Also this term in the AI world ‘x is solved’ i.e. music is solved… is a weird phrase when considering an invention used to give meaning to humans + their times. We gotta be careful

[@RabidDisposition]

How so? AI creating art won't stop humans from continuing to create. If you need to make money from your art for it to have meaning, it never had meaning to begin with.

[@wgxyz]

I think it depends who you are and your specific circumstances. Just because people make money from their art doesn’t mean it doesn’t have meaning. Everyone has to make a living. My point is really that if we are creating these tools, also be creating the support network for the disruption simultaneously. Btw I’m in general FOR expanding human creativity which these tools will obviously do.

[@RabidDisposition]

@@wgxyz making money for art doesn't invalidate the art's meaning, but NEEDING to make a living from your art for it to have meaning, makes the art meaningless.

[@yumeyuki1944]

I'm a musician. To me it's very exciting. There are some parts of music making process which are tedious. I think within next decade or few, AI won't be good enough to completely replace artists, but I think it will be a very useful tool (just like CGI in movies is). That being said, I think AI generated content is inevitable at this point. I think in the future you'll be able to enter a prompt, and see a randomly generated movie just like our brains generate dreams. Scary and fun times are comming.

[@guulish]

I’m so excited for the app on my phone that reads my face, smart watch biometrics, and mic that reads my mood and matches it too my taste for the perfect music.

[@hermannschmidt9788]

Considering how monotonous, stereotypical and boring contemporary pop music can be, I don't mind at all replacing it with AI. It's an endless repetition of familiar patterns, anyway. Real hand made music with emotion and competence is a whole different story.

[@cdeford]

They'll soon be writing full songs. The fourth one was the most impressive, that could easily have been written by a human.

[@jamilsuriel4406]

Humanity is finished.

[@hannahriley5560]

Investing in crypto now should be in every wise individuals list, in some months time you'll be ecstatic with the decision you made today

[@elliotalexander5624]

After watching so many KZbin tutorial videos about trading I was still making losses untill Akilina fan started managing my investment now, I make $6,800 weekly. God bless Akilina Nosova . She has been a blessing to my family.

[@mohammedadvik5254]

Please how can I contact Akilina Nosova fan? I really like what she has done for you, and I also want to benefit from it.

[@elliotalexander5624]

She's active on tele_gram

[@elliotalexander5624]

@Akilinanosova

[@mohammedadvik5254]

Thanks 😊

[@ZooxMaze]

@9:44 No 'bout-a-doubt-it... that's a 'flute' alright !! And, beautifully played too ! Totally, 'natural' :)

[@drvren030]

14:45 yeah I'd definitely wake up to that 😂 perfect wake up call

[@byujkt3822]

Oh, nice another person losing jobs and human become useless 😩 end of the day we got everything we want easily, let robot making money, and done the jobs. Happy ending for human 😁

[@norwegianblue2017]

Well it doesn't sound any worse than most modern music. It's just as soulless and formulaic.

[@aurasensor]

I may mention we had fully computer generated music in the 80s already. I think Chip 2000 it was called, or something like that.

[@drvren030]

Remember the good old days we were hating on autotune cause we didn't know better....yeah this is a million times worse. I just hope people feel that it's wrong they're not listening to an actual human voice behind that or genuine human effort behind that.

[@Bamazon1990]

We’re dancing around the final results: photorealistic animated shorts with voices sounds effects and music, ie movies made from prompts. Then wtf do we do

[@OmegaMouse]

LOL... i love how Google is stealing everyone's art and music and then reselling it back to the public as "A.I." generated. Those evil geniuses.

[@shagrat47]

What I love about this, is the prospect of thousands of copyright claims, breaking the legal system of the music industry, when AI basically creates every single possible song structure, melody etc. Sueing an AI that definitely did not "steal" anything in it's creative process can be interesting... 😂😂😂

[@krhodessch]

But, I can sue the creator behind what was fed into the bot. I would strongly recommend investing in the time in Shazaam before releasing this stuff.....before you get sued.

[@HikingWithCooper]

Wow, so much better than the music gen you did a recent video on. Hopefully Google’s AI comes out some day.

[@boomieboo]

Hal, give me Radiohead-style percussion and chords with a DeadMau5 electronic synth and melody. Then drop into an emotional interlude of isolated sweeping symphonic strings eventually building back into the main chorus and verses of the song consisting of all the aforementioned. If I get what I imagine out of a prompt like this, my producing dreams may finally be possible.

[@nocastus]

Gotta say, as a musician, I’m less impressed than I expected given the amazing output of ChatGPT and Stable Diffusion. First off, everything has a sort of fuzzy, out of focus texture, nothing like the clarity I’d expect from 24 bit 48 kHz. The rhythms are somehow off kilter too - there’s no sense of groove or intentionality, and this is before we start on the vocal problems. I’m sure AI will be able to do this soon, and I’m looking forward to asking it for the “Lost Beatles Album from 1967”, or “Check out this Spotify playlist and make something that fits”. But we’re nowhere near that yet.

[@TheGreatestJuJu]

*KZbin Creators are drooling…* unlimited original copyright free music for videos. Game changer!

[@AlexXanderMarketing]

Crazy good songs. Even the ones with the weird vocals were pretty on-point.

[@85pphoenix]

"Last night before I went to bed my jaw dropped to the floor" Wasn't totally sure where you were going with that 🤣😲🌭 Keep up the sweet content Matt

[@avi7278]

bruh