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:

-------------------------------------------------

▼ Extra Links of Interest:

-------------------------------------------------

Thanks for watching Matt Video Productions! I make all sorts of videos here on Youtube! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!

All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.

-------------------------------------------------

[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

[nickbarton]

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

[ofulgor]

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

[limelorax]

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

[joebender]

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."

[arnelilleseter]

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.

[avi]

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.

[johnnywoodstock]

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

[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! 😃👍😒🙄

[matthiasheymann]

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

[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.

[burkhardstackelberg]

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.

[KaiValentino-rw]

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!

[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.

[supermalavox]

4:31 Would a reggaeton song that makes you feel lost in space be called "De-space-ito"? #HorriblePun!

[cesar]

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

[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.

[Vansafe]

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.

[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. 🤯

[Je-Lia]

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