JPF Home Page Forums General Messages General Message Forum New Music Creation Tool Changes Everything

I tried one, it spit out a number of versions.

It didn't understand my structure but it's not terrible in some ways I'll share it soon.

Not better than mine but it offers some glimpses

It seems to favor straight up and down meters and feels. Bit sing songy

Try guiding it more toward the meter and feel you want in your prompt. Think as if you are talking to "music genie" who can only give you what you ask it for, and then, not too very specifically, unless YOU get really specific.

BTW, makes me happy to know you're having fun with it!

I’m not against AI in music. And non musicians could use this to offer a birthday song to a loved one… I haven’t looked at who owns Udio, but I bet it’s not some small potatoes company. Just looked. It’s founded by three guys formerly with Google AI.

Ok here is a song I wrote. Or just the lyrics entered for this experiment. I typed in the lyrics structure correct, but it decided to jump to whoever it wanted to.

It gives short take and never really gets too far into before it ends.

Pros? An instantaneous demo of your idea
Different variations and styles?

Cons? No dynamics no intros or motifs. Only decent melodies

It can lead to other ideas which I've already considered.

Not bad, but limited

Oh here are the versions I could hear potential in

https://www.soundclick.com/fdmuzik/?content=songs

Notice how my lyric sometimes gets cut up And moved around

She Fell Down

In my dream she was there
A hard angel in blue
At a gala somewhere
She was passin on through

She attracted the light
Like a magazine page
When the moment arrived
She walked up on a stage

And she fell down....
Yeah she fell down
I really wish shed fall down

Wasn't sure what I'd seen
Shook the cobwebs and strange
Things don't happen like this
Not to her unexplained

And she fell down
Yeah, she fell down
I really wish shéd fall down
I really wish shed fall down
I really wish shed fall to the ground
I'd pick her up and kiss her
If she fell dowm

If we're here with a plan
She was out here for man
Life comes easy to her
Like the palm of your hand

Now, she's only a face
Didn't know her too well
Through the city I'd chase
Only hope if she fell

And she fell down
Yeah, she fell down
I really wish shéd fall down
I really wish shed fall down
I really wish shed fall to the ground
I'd pick her up and kiss her
If she fell dowm

I like it!

The "jumping around" might occur...and I'm just guessing...might be cuz you asked it to sing too much lyric for the 32 second sample.

I set a Wordsworth poem and only fed it a verse at a time, and it gave me back verbatim, the words I fed it, then extended it w/the next part of the lyric, repeat, rinse...

So maybe just feed it less of the entire lyric, then add the next part of your lyric using the "extend" feature, and only write, in the "custom lyrics box" the NEXT part of the lyric that you want sung? Worked for me.

FD. Look inside the work box, you can add "intros," "outros" 32 sec parts before or after your first part, all using the "extend" button on the top right of the work box:

1) right click on the sample you want to extend

2)you will see Add Intro Before, Add Section Before, Add Section After, Add Outro
After

3) left click the type of extension to your sample that you want to highlight that type of extension

4) modify your prompt to reflect what you want in the next part

5) click on the "custom lyric" bullet to highlight it, and enter in the part of your lyric you want in the forthcoming extension

6) left click the "extend" button on the top tight and wait a few minutes.

7) choose between the two offerings or start over if you like neither...

8) repeat, rinse, adding an intro and outro, etc......

Dom

That is unbelievably good for click, wait a few seconds and go.
The musical interpretation of the lyrics is good.
The vocals and music are convincing.

And it gets better from here.

Yeah these are some really bright folks who partnered w/Adobe for the musical training data. AI music still has one final obstacle and it's holding continuity over the course of a song, but I fed it Wordsworth and got consistent structure, music, and performances, so Udio has seemingly tackled this problem, too.

Here is something I put together yesterday, using William Wordsworth's "I Wandered Lonely As A Cloud" as the lyric.

This creates very strong prosody, as does most classical poetry for it's strong, pronounced meter.

Many extensions later I had an unwieldy beast which I then tamed in FL Studio.

This is alt universe John Lennon type of singer, joins the Moody Blues, who after "Days of Future Past" decide to do another orchestral outing:

https://soundcloud.com/michael-j-za...ium=text&utm_campaign=social_sharing

You might find it bombastic, but honestly, I could not do a better job writing a composition to the poem. It seems perfect, to me, as a composition.

The recording less so. I could easily add a vocalist an octave up, in harmony w/the lower voice.

But this does show that Udio WILL stick to custom lyrics you feed it. It is word for word, the Wordsworth poem...

************************************************************************************

I Wandered Lonely as a Cloud
By William Wordsworth

I wandered lonely as a cloud
That floats on high o'er vales and hills,
When all at once I saw a crowd,
A host, of golden daffodils;
Beside the lake, beneath the trees,
Fluttering and dancing in the breeze.

Continuous as the stars that shine
And twinkle on the milky way,
They stretched in never-ending line
Along the margin of a bay:
Ten thousand saw I at a glance,
Tossing their heads in sprightly dance.

The waves beside them danced; but they
Out-did the sparkling waves in glee:
A poet could not but be gay,
In such a jocund company:
I gazed—and gazed—but little thought
What wealth the show to me had brought:

For oft, when on my couch I lie
In vacant or in pensive mood,
They flash upon that inward eye
Which is the bliss of solitude;
And then my heart with pleasure fills,
And dances with the daffodils.

After listening to several of the udio tracks I've noticed all were quantized, kind of a robotic feel. Even the vocals sound quantized. I'm not worried until this technology Advances to a more human feel.

John

I don't know, John...listen to this simple piano piece I "created" in Udio:

https://soundcloud.com/michael-j-za...ium=text&utm_campaign=social_sharing

Full of accelerandos, ritardandos, pauses...it's like the pianist is improvising, thinking about what to play next, leaning back in the chair to slow the tempo slightly, forward to speed it slightly.

Musicians and composers like you/me, have an advantage, here. I mean, I know how to get the music genie to give me an unquantized sound: I ask specifically for that in my written prompt...

I wasn't aware of the 30 second bit, probably the cause of it.

Thanks Marty , yeah I typed in the style I wanted, pretty detailed. one version made it sound like the cure. I like it but it's in the past.

But it is a much better reference than I thought

Mike...

I use Toontrack's EZKeys, EZBass, EZDrummer...ad all have "humanize" buttons that slightly randomize velocity and timing.
I suspect you know that.
Your average tiktoker will not realize a difference.

This stuff is not a snapshot. It is a moving target. What you may find lacking tonight, wont be in a month, year or very few years.
Every thing that has been put up in this thread has been impressive and at the same time, unsettling.

Agreed, it almost sounds like a cartoon character band. What it is is a decent reference.

If I wanted I could use some of the ideas, and play and sing myself.

It seems to be good at framing pop hooks.

I don't think it's anything more than a idea supporter on its own

But it's useful

This is so very true. Every drawback will certainly be addressed. But as I explained to John, I had no problem getting rubato, flowing, un-quantized sounding music simply by prompting Udio for exactly that.

Oh yes of course. Don't use those exact tools re:EZdrummer, etc., but imagine de-quant or humanize is a feature...

What if I was to redo this version

I have to admit the arrangement is perfect it just sounds 80s

https://soundclick.com/share.cfm?id=14763773

That could work.

There IS a "remix" button. If you like the arrangement, not the 80's sound, try "remix" and tell it "contemporary sound" in the prompt, or something to steer it out of the 80s...

Use my steps to create possible extensions, and don't overcrowd any 32 sec sample with too many words or it might not sing your custom lyric verbatim. You may need to experiment a little to get Udio to give you what you want. Modify your prompt accordingly. I find it fun, figuring out the magic words to get the results I want. I think you might, too.

I suppose a paid version gives you longer demos?

No, FD, you need to RIGHT CLICK the track and choose "extend" --you can get up to 5 minute songs!!!

Read my "step-by-step" further up this page.

Also, note, "remix" might be what you want, before you "extend" --if you want to take it out of the 80s...RIGHT CLICK on the track and choose "remix" from the menu, then modify your prompt, steering toward something more contemporary...

I'll keep an eye on this thread as long as you or others might need a little help figuring it out.

Do you think this would help me score chicks?


Ok yeah I'll have to fart around with jt more.


I can't help but say nooo wrong chord dude.

Id like to tell it what chord to play and wanna get in there myself.

But this could be enormous for our lyrics only people

LOL...I am slowly coming to understand your Jersey humor.

In the next thing I try, maybe tomorrow, I'm gonna see if I can prompt for specific chords. I have no idea if that will work, but imagine, sooner or later, it will...if not now, in future versions...

By the way, it's only free while in "beta" and that might mean from a couple weeks to a couple months, I have no idea...

There's no doubt in my mind that as we speak, that band in a box is working on getting ai in there package

Cause if you can get all ai has to offer with edit ability that's a slick combo

I just feel like I wanna be in the demo

It has a dreamy feeling to it. Did you write all the notes? Or just submit a phrase or so? I'm not fond of the piano sound on this. It would be interesting if you'd do a home version of this on your workstation for comparison.
Here's a piano track I composed in the style of George Gershwin. I don't see this technology doing something like this: http://schicksville.com/The%20Days%20of%20Gershwin.mp3

John,

My prompt was something like this: A slow, dreamy post-minimalist piece for piano, with an improvised feel using mostly broken minor chords, and thoughtful tempo ritardandos.

Even for someone like you who needs to absolutely NOT use AI in your recordings, what would be stopping you from prompting it to get piano pieces you could then play yourself?

The idea is, if nothing more, Udio can be something that can give you an idea that you would never have thought of on your own. You can't prompt for specific modern artists/composers, but what it does is it takes, say, "in the style of Max Richter" and rewrites that to something like "modern tonal post-minimalism with baroque note ornamentation" kinda thing.

You CAN prompt for older composers, as in this prompt: a piano piece that is a cross between Beethoven and Chopin, full of block chords but also right hand melodic runs, modulating between several related keys, with unexpected silences where one expects big downbeats" --another words, you can be really specific. Or even throw "word salad" at it and see what you get! The idea being that if you get something both surprising AND you like it, then what is to stop you from learning how to play it yourself? Then your publisher would be none the wiser. Using Udio as a compositional tool is probably what I will end up doing with it mostly, myself.

EDIT--it does the same thing with older composers, my bad. It will break "Beethoven" down into a written expression of his style though...I am quite surprised and assumed one could prompt for composers whose work is in the public domain, but no.

Very Gershwin-y indeed!

Yes, more "presence" in your virtual piano than what Udio created, but I DO like that in the Udio piece it did not mess up the sustain pedal though it was used all the way through.

Your piece was very nice, btw...

It does have a dreamy feel to it. Just doesn’t measure up to John’s Gershwin. And maybe the Udio team would agree for now. All very interesting. John mentioned that his sync. publisher (who places his music) would not allow any AI assist. I would suspect they’d want to stay away from this due to any copyright issues that might come up in the future. Michael: on the Wordsworth attempt, did you direct it to sound like the Moody Blues? It does a bit, but though the prosody might work okay with this, the music is kinda monotonous. Looks like you’re figuring this out along the way. Looking forward to more songs to hear how this evolves.

Yes, probably a copyright issue thing concerning AI usage w/John's publisher.

About the two piano piece, that's fair to say, and I agree somewhat, but also think comparing the two piano pieces is a little apples/oranges in that they have entirely different directives. IOW, John's piece isn't going for dreamy and mine isn't going for "early 20th century Cosmopolitan."

Didn't prompt for The Moody Blues. It was solely an experiment to see if I could get consistent form and music, and so "monotonous" is an appropriate response, since my main goal was consistency from start to end, since that is the last unconquered territory of music AI, and it truly did succeed in making the last three sections very similar, albeit too similar to not sound monotonous to some LOL...and that's okay...it successfully sang the entire lyric while keeping a coherent structure and that was the goal. Creating a masterpiece the first day I use ANY software is not something I am attempting, so thanks for understanding that...I save that for day 4...make that day 44.

I rather like the composition it created and will probably add a voice either up a third or an octave to give it more emotion.

BTW thanks for listening, Craig.

Interesting Michael. Have you presented Udio with lyrics to a tune you’ve already written to compare how the two sound?

No, but I might eventually try that. But feeding it an orphaned lyric, that's more likely in the near future. I have too many of those.

I'll use Lalai's online stem creation as well so as to have more control over adding effects to specific stems.

No but Dom has

I listened to She Fell Down earlier today but didn’t see your version Dom. Did you think it compared well? I know this is new. Seems Udio just started. Udio’s version of She Fell Down wasn’t too bad actually. I just think these Google guys with Udio will sell it to some big company for lots of bucks and use the money to find the next big (human) musical phenom.

This is insane

Here's a Joanie Mitchell version of my song

https://www.soundclick.com/share.cfm?id=14763870

Here's a smithereens version

https://www.soundclick.com/share.cfm?id=14763869

I think it's a brilliant idea to use a female artist, for various reasons. One being, there's less chance a listener could possibly (erroneously) misinterpret the singer's intention.

This is insane the first jazz song I ever tried to write.

I can't really play jazz but it translates

I just did the chorus and I cant figured out how to keep this killer chorus and now add the verses

https://www.soundclick.com/share.cfm?id=14763867
The slower you get
The faster the women walk by
That sweet lovin scent
Ain't no good reason why
But somehow it happens
as in the blink of an eye
The slower you get
The faster the women Walk by

Type, click and presto...

Incredible.

I wanna hear other people doing this. I believe the structure and prosody matter, so does flow and of course something to say

Id like to hear others before I commit myself to this, cause it is amazing me

What I've observed so far is that it dives into the past to find sound, style, vocal quality etc

That's it's limitation right now.

Everybody using is going to sound basically the same, while the songs may not be

That's kinda why I wanna hear others doing their songs with it

Dom, you can add 32 sec samples "before" or "after" your initial 32 sec sample by "right clicking" the sample name on the left hand side of your "my creations" page (or the tiny box to its left) to get to your main menu....then choose "extend" ...here's the whole procedure:

1) right click on the sample you want to extend

2)you will see Add Intro Before, Add Section Before, Add Section After, Add Outro
After

3) left click the type of extension to your sample that you want by highlighting that type of extension

4) modify your prompt to reflect what you want in the next part

5) click on the "custom lyric" bullet to highlight it, and enter in the part of your lyric you want in the forthcoming extension

6) left click the "extend" button on the top right and wait a few minutes.

7) choose between the two offerings or start over if you like neither...

8) repeat, rinse, adding an intro and outro, etc......

Thanks Mike

30 seconds before and after won't be enough. I tried doing it over and it gives me a big band version

I want that chorus that's what I always wanted from that song

You can extend up to a little over 5 minutes. Just keep repeating the process using the part of the lyric that fits the particular 30 sec sample you're working on.

Extending doesty seem to work it says error

Hmmm...that's weird, I haven't run into not letting me "extend" my track. Try it again. If it still doesn't work, try to troubleshoot:try extending a different track. If you can, then the track you get "error" on has a problem. But if you get "error" on the new track, it's likely you are perhaps not in the right place, left clicking the correct "extend" button...

Dom, here's a short video showing you, to make sure you're doing it right:



Note, I am

1) right clicking on the name of the song to open the main menu

2) then I am left clicking "extend track" in that menu

3)then in the big pop up box I can mess with the prompt, add the correct lyrics, etc., then left click "extend" which you see in the top right of the big rectangular box

It's ,430 and I'm still doing this

Obsessed...

I tried the song sy least 25 times and I can't get the feel I first got.

I'm frankly really pissed off now

Guess I should try a different tune

I think though that this stuff sounds like a real album, but in a way that seems off for some reason. And it tends to like melodies that don't jump around much

I think you will realize after a while that most of this stuff has been done already and not really original

But, it's fun and cool and I think great as a songwriter tool

Bottomline, I think it’s the beginning of the end to originality and creativity by letting this application do your thinking for you. Which it basically is doing. Of course, I'm just an old codger set in my ways - ha, ha.

Carry on, John

Just imagine if an AI created song won the song of the year. Something created by a soulless machine competing against a human being, especially if it was a Gospel song. I'm about to quit writing anyway, this would be the push I need to throw up my hands and say, what am I knocking myself out for.

I like to think that I have an aptitude for writing. It's arguable that I dont. But I cling to the delusion.
My personally recognized shortcomings are melody creation and melody loveliness. I strive for both, but end up... limited by a propensity to fall into the same familiar chord ruts and a limited vocal range.

I am hopeful that I can get some help with that from this AI.
I don't expect this to spit me out a good-to-go instrumented tune that I like (in 2024) but it may give me a chord progression/ vocal melody that I like and can work up and would have never arrived at entirely on my own.

I plan to see.
Stated another way...I like to write alone. It's therapeutic and puts me in the moment, but this may be the co-writer that I was looking for.
(until the technology derives all the information that it can from me and people like me (betas) and leaves us discarded and unnecessary. )
It's is likely that we are the employees who are training our replacements before being laid-off.

Martin