Udio's New Feature is a Gamechanger for AI Music

Theoretically Media
6 Jun 202414:07

Summary

TLDRIn this video, the host explores the new AI-generated music feature released by Udo, which allows users to upload their own music and extend it with AI. The host tests the feature with various prompts and music styles, demonstrating its ability to generate unique tracks and lyrics. They also experiment with AI voice models, creating a John Mayer-like vocal track, and discuss the potential for further experimentation and the limitations of the current technology, such as the inability to separate individual tracks.

Takeaways

  • 🚀 AI-generated music has made significant advancements with the release of Udo's new upload feature.
  • 🔗 There's a tutorial available for understanding how Udo works, linked in the description.
  • 💰 The new 'Extend Your Own Audio' feature is available for paying users, not just those on the Pro Plan.
  • 🎹 The feature allows users to upload their own music and extend it with AI-generated content.
  • 🎼 The script includes a demonstration of extending a solo piano piece performed by a world-class musician.
  • 🎶 Udo captures the characteristics of the original recording and matches them in the AI-generated output.
  • 🔄 There are limitations to the feature, which are discussed in the script.
  • 🔁 The script explores experimenting with AI by stacking different AI-generated tracks to create unique sounds.
  • 🎤 Udo can generate music with user-provided vocals, but there are challenges in getting it to sound like a specific artist.
  • 🎧 The script includes an example of using AI to generate music in the style of John Mayer using a vocal sample.
  • 🛠️ The potential of AI in music creation is highlighted, including the ability to inspire new ideas and techniques.

Q & A

  • What is the main topic discussed in the video script?

    -The main topic discussed in the video script is the new upload feature of an AI music generation tool called 'udio', and how it has advanced in creating music based on user-uploaded audio.

  • What is the significance of the 'extend your own audio' feature in uadio?

    -The 'extend your own audio' feature in uadio allows users to upload their own music and have the AI generate new music based on the uploaded audio, which is a significant leap in personalizing AI-generated music.

  • Is the new feature only available to users on the Pro Plan of uadio?

    -No, the new feature is not exclusive to the Pro Plan. The script mentions that the user only has the standard plan and still has access to the new feature.

  • What is the basic concept of the new uadio feature?

    -The basic concept of the new uadio feature is to allow users to upload a short snippet of their own music and then provide prompts to uadio to generate an extension or continuation of that music in a style of their choosing.

  • What is the role of the piano piece performed by a world-class musician in the script?

    -The piano piece serves as an example to demonstrate the capabilities of the new uadio feature, showing how a short snippet of music can be uploaded and extended by the AI.

  • What is the name of the collaborative track mentioned in the script?

    -The name of the collaborative track mentioned in the script is 'Rain on Glass'.

  • What is the significance of the four-chord loop in the script?

    -The four-chord loop is used to show how uadio can take a simple musical idea and generate an entirely new piece of music in different styles based on the user's prompts.

  • What is the result of combining multiple AI-generated tracks from uadio?

    -Combining multiple AI-generated tracks can result in a more complex and layered piece of music, as demonstrated in the script where the combination of tracks created a sound similar to a theme song for an HBO crime drama.

  • What is the process of using AI to generate music with vocals?

    -The process involves using a tool like Aon to generate a vocal track from a sample, and then using uadio to generate instrumental tracks based on the vocal sample and given prompts.

  • What challenges did the user encounter when trying to generate music with the voice of John Mayer?

    -The user encountered challenges with generating lyrics that matched the vocal melody and the style of John Mayer. The initial results sounded unnatural, and the user had to use additional tools like chat GPT to refine the lyrics.

  • What is the potential workaround for generating music with specific vocals like John Mayer's?

    -A potential workaround suggested in the script is to generate an a cappella version of the vocals first and then generate an instrumental track based on that, merging the two to create a complete song.

  • What is the user's opinion on the future of AI-generated music tools like uadio?

    -The user believes that AI tools like uadio shine in their ability to inspire and assist in the creative process, providing new ideas and directions that the user might not have thought of otherwise.

Outlines

00:00

🎼 Exploring AI-Generated Music with Udo's New Feature

The video script introduces a significant advancement in AI-generated music with Udo's new upload feature. The narrator spent time experimenting with the feature, pushing it to its limits. It's clarified that the feature is available to paying users, not just those on the Pro Plan. The feature allows users to upload their own music and create extensions before or after their song. A demonstration is given using a solo piano piece, showcasing how Udo captures the characteristics of the original recording. The narrator also discusses limitations and the potential for further exploration with the new model.

05:00

🎹 Experimenting with AI Music Generation and Kit Bashing

The script continues with the narrator's exploration of Udo's AI music generation capabilities. Using a four-chord loop from a collaborative track, the narrator demonstrates how Udo can extend and transform the music into different genres, such as indie rock electronic and downtempo. The process of stacking different AI-generated tracks to create a complex composition is discussed, highlighting the creative potential of combining multiple AI outputs. The narrator also touches on the topic of generating music with AI using a user's own voice, referencing a previous video and the use of a tool called Aon to create a John Mayer-like vocal performance.

10:02

🎤 AI Music Generation with Custom Vocals and Guitar

In the final paragraph, the narrator delves into the use of AI for creating music with custom vocals and guitar parts. They experiment with using a John Mayer vocal sample and Udo to generate a blues rock track, encountering some challenges with the initial results. The narrator then uses Chat GPT to refine the lyrics and re-inputs them into Udo for a better outcome. The script concludes with a humorous anecdote about creating an 'AI John Mayer' performance at a fictional venue, reflecting on the creative possibilities AI tools offer for generating unique musical ideas and the current limitations regarding vocal generation.

Mindmap

Keywords

💡AI Generated Music

AI Generated Music refers to the creation of musical compositions using artificial intelligence algorithms. In the video, the host discusses the advancements in this field, particularly with the release of a new feature by 'udio'. The script mentions how the host experimented with this technology, pushing it to create unique musical pieces, which is central to the theme of exploring AI's role in music creation.

💡udio

Udio is the name of the AI music generator discussed in the script. It is highlighted as having a new feature that allows users to upload their own music and generate extensions or new compositions based on it. The host uses 'udio to demonstrate the capabilities of AI in music production, showing how it can extend and transform existing musical pieces.

💡Music Extension

Music Extension in the context of the video refers to the process of adding new sections or parts to an existing piece of music. The new feature of 'udio allows users to create extensions before or after a song, effectively expanding the musical composition. The host demonstrates this by uploading a solo piano piece and extending it with a classical prompt.

💡Paying User

A paying user is someone who has a subscription or has made a payment to access premium features of a service. In the script, it is mentioned that the new 'udio feature is available to paying users, indicating a business model where certain functionalities are exclusive to those who have financially supported the service.

💡Solo Piano Piece

A solo piano piece is a musical composition performed by a single pianist without any accompaniment. In the video, the host uploads a solo piano piece to 'udio and uses it to demonstrate the AI's ability to extend and generate music based on the uploaded piece, showcasing the AI's capacity to understand and build upon complex musical structures.

💡Kit Bashing

Kit bashing is a term used in music production that refers to the process of creatively combining different elements or 'kits' of sounds to create new and unique compositions. The host mentions 'kit bashing' in the context of experimenting with 'udio, pushing its capabilities to see how it can blend and transform various musical elements.

💡Indie Rock

Indie Rock is a genre of music that originated from the independent music scene and is characterized by its DIY ethos and often alternative sound. In the script, the host uses 'Indie Rock' as one of the prompts for 'udio to generate a new musical piece, indicating the AI's ability to understand and produce music within specific genres.

💡DAW (Digital Audio Workstation)

A Digital Audio Workstation (DAW) is a software application used for recording, editing, and producing audio files. The host mentions using Ableton, a popular DAW, to stack and combine different AI-generated tracks, demonstrating how DAWs can be used in conjunction with AI tools to create complex musical compositions.

💡Aon

Aon is a tool mentioned in the script that generates music from a trained voice input. The host uses Aon to transform a vocal hook sample into the style of John Mayer, illustrating how AI can be used to emulate the vocal characteristics of specific artists, which is a significant aspect of the video's exploration of AI in music.

💡Lyrics Generation

Lyrics Generation refers to the process of creating song lyrics, often using AI algorithms. In the video, the host discusses the AI's ability to generate lyrics, as seen when 'udio creates lyrics for an indie pop song. The script highlights the potential and challenges of AI in crafting meaningful and rhythmically appropriate lyrics.

Highlights

AI-generated music has made significant advancements with the release of Udo's new upload feature.

The new feature allows users to upload their own music, enhancing the extension feature of Udo.

The feature is available to both standard and Pro Plan users, contrary to previous reports.

Udo's new model captures the characteristics of the original recording for extension.

A basic demonstration showcases a solo piano piece transformed into a classical extension.

Experimentation with a four-chord loop from the song 'Rain on Glass' resulted in unique AI-generated tracks.

The AI took the chords and arpeggiated them, creating a new genre-appropriate track.

Stacking different AI-generated tracks can create complex and atmospheric compositions.

AI can generate lyrics and entire songs from simple chord loops.

Experimentation with vocal samples and AI models like Aon can mimic artists' styles.

A vocal hook sample was transformed into a John Mayer-like rendition using AI.

Udo's advanced features allow for customization of generation quality and lyric strength.

AI-generated music can struggle with generating lyrics that match the original vocal melody.

Chat GPT can be used to refine AI-generated lyrics for better coherence.

Combining AI-generated vocals and instruments can produce unexpected and creative results.

The potential workaround for generating a complete song with vocals and instruments involves using separate AI models for each.

AI tools like Udo are expected to evolve, with stem separation being a highly anticipated feature.

The comparison between Udo and other AI music generators like Sunno is noted, with anticipation for their upcoming features.

Transcripts

play00:00

so AI generated music has taken a

play00:02

massive leap forward with the release of

play00:04

udo's new upload feature today we're

play00:06

going to dive into this new feature and

play00:08

I spent some time this morning like

play00:09

really digging into it I ended up

play00:11

pushing things in some pretty

play00:12

interesting directions using some kit

play00:14

bashing trying to break it basically uh

play00:17

yeah you're definitely going to want to

play00:18

check this out okay let's dive in so

play00:21

first off if you actually need a walkr

play00:23

of how UD works there is a tutorial that

play00:25

I did earlier that is linked down below

play00:28

secondly this new extend your own audio

play00:30

is only available if you are a paying

play00:32

user I think it was previously reported

play00:34

that you had to be on the Pro Plan in

play00:36

order to use this that is not true I'm

play00:38

actually only on the standard plan and I

play00:40

have access to it so this new feature is

play00:42

basically a spanked up version of their

play00:44

extension feature which allowed you to

play00:47

you know create extensions after your

play00:48

song or before your song at an intro or

play00:51

at an outro the difference this time is

play00:53

that you can upload your own music so

play00:57

kicking off with a very kind of basic

play00:58

version of this uh here is a solo piano

play01:01

piece performed by a worldclass musician

play01:04

down in Studio A I'll tell you more

play01:06

about that in a second uh let's go ahead

play01:08

and

play01:09

[Music]

play01:12

listen and that is how you tickle the

play01:15

ivories now that's only 3 seconds you'll

play01:17

actually hear the entire piece in a

play01:19

video next week I'm going to hold off on

play01:21

telling you who the surprise guest is

play01:23

but if you've been following the channel

play01:24

there's a pretty good chance you've

play01:26

figured it out anyhow uploading that

play01:28

short snippet and then giving udio The

play01:30

Prompt classical piano contemporary

play01:32

classical just keeping it very simple uh

play01:35

and we're going to use the ad section

play01:36

after we end up with

play01:41

[Applause]

play01:42

[Music]

play01:52

this and that is pretty impressive now

play01:55

admittedly I have a bit of a bias

play01:56

considering that I have heard the

play01:58

original record Rec ing uh and you will

play02:01

too next week don't forget to hit that

play02:02

subscribe button but I think what's most

play02:04

interesting and fascinating to me is the

play02:06

fact that this new udio model actually

play02:08

takes the characteristics of the

play02:10

recording and matches to that I haven't

play02:13

had a chance yet to EQ and sweeten that

play02:16

recording so what you heard was pretty

play02:18

much just you know microphone recording

play02:20

and udio really did capture those

play02:23

characteristics now there are some

play02:25

limitations on that front and we'll talk

play02:26

about that in just a little bit so

play02:28

here's where things get really

play02:30

interesting so I ended up taking a four

play02:32

chord Loop that I recorded for a song

play02:34

called rain on glass this was a

play02:36

collaborative track that I did with the

play02:38

UK producer named palid it's kind of one

play02:40

of those like sleepy ambient Loi tracks

play02:43

it's up on Spotify Apple music and all

play02:44

those other streaming services if you

play02:46

want to listen to it hey look uh 11,000

play02:48

listens meaning that we made about 2

play02:51

cents off of that streaming royalties

play02:53

are a whole other rant video we're not

play02:54

going to get into that now so very

play02:56

briefly here are the chords

play03:01

[Music]

play03:06

those four chords more or less Loop and

play03:08

we just put some ornamentation around it

play03:09

now let's see what udio does with it so

play03:12

running it through as an instrumental

play03:13

track and giving it an extension uh we

play03:16

ended up with neon Drifter the prompts

play03:17

here were indie rock electronic

play03:19

downtempo and left field uh let's take a

play03:22

listen to what we got

play03:28

[Music]

play03:33

so not bad definitely within genre and

play03:35

what's interesting is that it ended up

play03:36

taking my chords and kind of

play03:38

arpeggiating them now udio does generate

play03:41

two tracks per you know generation the

play03:44

second one I wasn't necessarily

play03:45

incredibly Blown Away with at first

play03:48

[Music]

play04:01

so in this case it did end up using the

play04:02

original chords but then adding in a lot

play04:04

of like weird atmospheric stuff and that

play04:07

got me thinking well what would happen

play04:08

if we started stacking these things so I

play04:10

brought them into a DA digital audio

play04:12

workstation uh I use Ableton you can

play04:14

pretty much use any audio editor that

play04:16

you want to accomplish this so on this

play04:19

track I just have you know those

play04:20

original guitars that we supplied udio

play04:22

with the second track was our

play04:24

atmospheric e one and the third track

play04:26

was the generation that had the heavier

play04:27

bass and drums but when we turn turn all

play04:30

three of them on it suddenly sounds like

play04:31

it's the theme song to a new HBO crime

play04:34

drama

play04:37

[Music]

play05:00

now while that can use some editing I

play05:01

think one of the coolest things about

play05:03

all of this is there's no such thing as

play05:05

a bad or off generation in fact if

play05:07

anything it just kind of makes you want

play05:09

to experiment more for example you know

play05:11

on the channel I definitely tend to lean

play05:13

more towards instrumental stuff although

play05:16

we do have some pretty crazy vocal stuff

play05:18

coming up in just a minute so while

play05:20

experimenting I decided to let udio take

play05:22

that four chord loop again uh I just

play05:24

changed out the prompt to Indie pop and

play05:26

let udio write lyrics and we ended up

play05:28

with this me what's your heart's desire

play05:32

make it clear don't let it hide I just

play05:36

got to see what's your favorite

play05:41

Skyline what you wish a star tell me am

play05:46

I on your mind do we

play05:50

align is this

play05:54

fine

play05:57

oh I just got to

play06:02

[Music]

play06:03

see so yeah that's pretty much a song

play06:06

and we can kind of just keep generating

play06:08

iterating cutting and looping and you

play06:10

know very easily come up with a 3 to 5

play06:13

minute long original song and that's

play06:16

without any other stem separation or

play06:18

fancy tricks although obviously if you

play06:21

have fancy tricks you could probably go

play06:22

even crazier with this which does lead

play06:25

us into crazy kit bashing territory so

play06:27

one of the questions that's probably

play06:29

most asked every time that I do a story

play06:32

on an AI music generator is can I upload

play06:34

my own voice and have it create you know

play06:37

a song and the answer to that is yes

play06:39

sort of we're going to go into a really

play06:41

weird Direction here caveat I am not

play06:43

singing you do not want to hear me sing

play06:45

instead we're going to revisit an old

play06:47

video that I did and one that nobody

play06:49

watched cuz it was on AI girlfriends in

play06:51

the back end of that video I ended up

play06:52

taking a look at a tool called ampon

play06:55

This is a model that will generate from

play06:56

a trained voice off of a voice input so

play07:00

with Aon I ended up taking a vocal hook

play07:02

sample and turning it into John Mayer

play07:04

quick side note despite being a white

play07:05

dude in his 40s that plays guitar I'm

play07:07

actually not the biggest John Mayor fan

play07:09

nothing against him I think he's an

play07:10

incredible guitar player I I've tried to

play07:12

learn neon numerous times it's very hard

play07:15

but I do like his stuff with the trio

play07:16

and actually have seen him a couple of

play07:18

times with the Grateful Dead maybe I do

play07:19

like John Mayor damn you mayor so anyhow

play07:22

taking this vocal sample is it really

play07:24

love is it really love or am I just

play07:28

falling for the things they could call

play07:31

it l but in you I trust let's forget

play07:34

about the world cuz you're where I want

play07:36

to be and running it through Aon with

play07:38

the John Mayor model we end up with this

play07:40

is it is or am

play07:44

iing for the same things they could call

play07:47

it l but in you I trust let's forget

play07:51

about the world cuz you're where I want

play07:52

to be yeah Anana is kind of crazy I'm

play07:55

glad you guys are catching it now cuz

play07:56

again nobody saw it in the AI girlfriend

play07:58

video so loading that John Mayor sample

play08:00

into udio and giving it you know prompts

play08:03

like blues rock and uh let's do like

play08:06

drums bass guitar and then not giving it

play08:08

any custom lyrics Just To See what'll

play08:10

happen uh I do want to just point out

play08:12

something here if you actually prompt

play08:13

for John Mayer as well you'll actually

play08:15

end up with a little no no prompt uh it

play08:17

won't actually give you John Mayer

play08:19

instead it will replace it out with

play08:20

things like male vocalist Americana

play08:22

singer songwriter North American music

play08:25

all the country I guess uh and mellow I

play08:27

should add under the advanced features

play08:29

you have things like generation quality

play08:31

uh you should probably just crank that

play08:32

up to ultra uh the lyrics strength as

play08:34

well the higher that you kick it the

play08:37

possibly more accurate the anunciation

play08:39

will be but may not sound as rhythmic

play08:43

and natural um we're just going to leave

play08:45

that here and then for our prompt

play08:46

strength we're just going to leave that

play08:47

I'm going to take that up to 66 why not

play08:50

uh let's run this and see what we get

play08:51

the initial results were not that great

play08:53

in fact if anything it just kind of ends

play08:54

up sounding like John mayor's having a

play08:56

stroke let's take a listen to

play08:57

stratospheric strut here saying I'm for

play09:01

the same

play09:09

things so interestingly it still sounds

play09:11

like John Mayer and it definitely has

play09:13

like this kind of rhythmic flow to it it

play09:15

kind of sounds a little bit like a demo

play09:16

version of a song that he might be

play09:18

writing that he hasn't figured out the

play09:20

lyrics for I do have a solve for this

play09:21

we're going to talk about that in one

play09:22

second but I just before we did I I have

play09:24

to play you Continuum drift because uh

play09:26

this one just makes me laugh

play09:37

John if you need do do do backups on

play09:39

your next album give me a call so really

play09:41

I think the problem came down to the

play09:42

fact that udio just really wasn't able

play09:44

to generate lyrics at that front end it

play09:46

was just it did have vocal Melody so

play09:48

what I decided to do was take the lyrics

play09:51

that uh we initially had bring them over

play09:53

to chat GPT and listen I Know lyrics and

play09:56

chat GPT look this is a YouTube video

play09:58

I'm not going to pour my heart out like

play10:00

writing lyrics for this so bringing that

play10:02

back over to udio we ended up with

play10:03

something that actually came out pretty

play10:05

good the fact we could take it

play10:09

slow we could take it

play10:12

slow every moment with

play10:15

you feels like a

play10:18

dream but life so fall love

play10:25

to so interestingly it does fumble in

play10:28

the transition between our initial set

play10:30

of lyrics and our sample lyrics There's

play10:32

kind of like this jumble of words there

play10:34

in the center but I think if you wanted

play10:36

to use this you could probably just edit

play10:38

that section out now I will say that the

play10:39

downside to all of this is that I was

play10:41

not able to get a ban behind John Mayer

play10:45

no matter how much I prompt or remixed

play10:48

or extended backwards or forwards but

play10:50

when I initially created that John Mayor

play10:52

voice in the AI girlfriend video my AI

play10:55

girlfriend is a huge fan of John Mayor

play10:57

by the way I decided to have a little

play10:58

fun with it so I decided recorded track

play11:00

underneath it let's check that out real

play11:06

quick for the sa they could call it l

play11:10

but you trust let's forget about the

play11:13

world cuz you're want to be thank you

play11:15

for indulging me there AI John Mayer and

play11:17

I are playing the Dairy Queen this

play11:19

weekend it's boob so bringing that track

play11:21

into udio the results were both

play11:23

absolutely stunning and mid at the same

play11:26

time it was really weird uh let's take a

play11:28

listen to in my arm

play11:29

in particular note how much better a

play11:31

guitar player udio is than I

play11:33

[Music]

play11:52

am we could take it slow we could take

play11:55

it slow every moment with you feels like

play11:58

a dream when you're in my arms in my

play12:02

arms Nothing Else Matters it's just you

play12:05

and me is there

play12:08

any so yeah it's actually pretty cool I

play12:10

mean the biggest thing to notice is that

play12:11

Jon's voice obviously is very different

play12:15

from the sample that we provided but man

play12:17

that udio guitar player definitely gave

play12:19

me some pretty cool ideas in another

play12:21

generation and I'm just going to play a

play12:23

quick second of this uh udio just

play12:24

decided to go full John frante with my

play12:27

guitar John frante the guitar player

play12:30

from the Red Hot Chili Peppers

play12:32

straightup guitar genius uh he is very

play12:35

much a guitar players

play12:39

[Music]

play12:49

[Applause]

play12:51

guitarist we could take slow this is

play12:54

where I think that udio and AI Tools in

play12:56

general really shine I probably never

play12:58

would thought of going full like

play13:00

Hendrick's feedback for that opening but

play13:02

you know now that I've heard it I could

play13:04

either you know maybe use it if I wanted

play13:05

to cheat it or I could go out and learn

play13:08

it and record it now as far as the John

play13:10

voice I don't really have a trick for

play13:12

that yet but I do think there would be a

play13:14

workaround by generating up an Acappella

play13:16

and then generating up a track after

play13:19

that and then merging them together I

play13:21

actually don't have time to try that out

play13:22

today let me know in the comments if you

play13:24

would like to see me give that a go and

play13:26

before anybody starts yelling at me in

play13:27

the comments yes I do do know that sunno

play13:30

has a very similar feature that is

play13:32

getting ready to release I re I mean

play13:34

what can I say I reached out toso I

play13:36

asked if I could try it out and you know

play13:38

I never heard back from them regardless

play13:40

I do think that sunno will be dropping

play13:42

that soon and then after this I think

play13:43

the only real major thing that we're

play13:46

going to be looking for is the ability

play13:48

to be able to separate out the

play13:50

individual tracks so I'm curious to see

play13:53

who gets to that one first until then I

play13:56

thank you for watching my name is Tim

Rate This

5.0 / 5 (0 votes)

Related Tags
AI MusicUdo FeatureAudio UploadMusic CreationPiano PieceClassical PianoIndie RockElectronic MusicAmbient LoopVocal GenerationLyrics Writing