This is the first such generator I've seen that not only makes coherent, realistic-sounding music -- Jukebox was awful in comparison, and even Suno has a weird synthy quality to the vocals -- but...
This is the first such generator I've seen that not only makes coherent, realistic-sounding music -- Jukebox was awful in comparison, and even Suno has a weird synthy quality to the vocals -- but also makes songs that are genuinely catchy.
Jesus. As a musician who wrote a few years ago about AI's ability (or lack thereof) to make art, this caught me wildly off-guard. woo reading comprehension
Jesus. As a musician who wrote a few years ago about AI's ability (or lack thereof) to make art, this caught me wildly off-guard.
What was your prompt for "Website On the Internet"? woo reading comprehension
I didn't make any of them, just pulled my favorites from the top charts, but the tags for that song list: Barbershop, Rhythmic, Happy, Melodic, Playful, Vocal group, Warm, 1950s Barbershop,...
I didn't make any of them, just pulled my favorites from the top charts, but the tags for that song list: Barbershop, Rhythmic, Happy, Melodic, Playful, Vocal group, Warm, 1950s Barbershop, Rhythmic, Happy, Melo...
I was playing with Suno to make a version of the New Zealand national anthem. I asked for polka as a joke, but it made something actually listenable, rather than a chore to suffer through prior to...
I was playing with Suno to make a version of the New Zealand national anthem.
I asked for polka as a joke, but it made something actually listenable, rather than a chore to suffer through prior to sports games, school assemblies, or state events.
The Māori words are butchered, but in the same way that most people would sing them.
The English verses (or at least the more well known second verse) begin with the same loudness and chorus that happens as people join in singing since that's the words they know.
“You Spilt a Coffee on My Dog” is a standout for me. It sounds like a Robyn song. I genuinely adore the lines “ain’t no mistake you dropped that brew / what my dog ever do to you?”. A lot of...
“You Spilt a Coffee on My Dog” is a standout for me. It sounds like a Robyn song.
I genuinely adore the lines “ain’t no mistake you dropped that brew / what my dog ever do to you?”.
A lot of AI-generated content has a sort of soulless uncanny-valley effect on me that’s off-putting, but this is great. I’m legitimately into it.
Um... wow. I've known about the advances in generative AI of course, but 'being aware' is a different experience from listening to music that has vocals I can't discern being computer generated...
Um... wow.
I've known about the advances in generative AI of course, but 'being aware' is a different experience from listening to music that has vocals I can't discern being computer generated while at the same time the music is actually good to my tastes.
Well. It's about time I personally got something nice from Skynet before... you know... Skynet does Skynet things.
I… Uhhh… wow My honest reaction I even asked it to sing a song about a city in Portugal, and I wrote my prompt in English, I didn’t expect it to actually deliver a portuguese song using Fado
I… Uhhh… wow
My honest reaction
I even asked it to sing a song about a city in Portugal, and I wrote my prompt in English, I didn’t expect it to actually deliver a portuguese song using Fado
This is awesome! The editing tools are really neat - you generate your song in increments of ~30s or less -- well, technically 30s each time, but you can trim out any parts of those 30s you don't...
This is awesome! The editing tools are really neat - you generate your song in increments of ~30s or less -- well, technically 30s each time, but you can trim out any parts of those 30s you don't want to keep -- and best of all imo, you can redo your prompt at each stage, to evolve the style of the song as you go while trusting that it'll still make the whole thing fit together in the end.
Here's my first take at a full-length track: an overdramatic symphonic metal/gothic rock song about being sad that your cat won't come over to cuddle you lol (I had fun with it)
There are a few options for that: The most direct way is exclusive to paid subscribers only ($10/mo USD for the basic plan), which gives you the option to "inpaint" over an already generated song...
There are a few options for that:
The most direct way is exclusive to paid subscribers only ($10/mo USD for the basic plan), which gives you the option to "inpaint" over an already generated song & selectively tweak specific sections (either for lyrics, to specify a different gender of vocalist, or other minor style tweaks);
Everyone can just directly supply a full set of lyrics for it to use in the first place - not exactly "fine" tuning, but hey); or
You can also copy-paste the lyrics it made up, then go to regenerate that section of the song altogether again, but specify the full set of lyrics as whatever you copy-pasted with your edits. By default, this will give you something that sounds different from the first time (since it's regenerating it) -- but, if you do your initial generation with a set seed (one of the options under advanced settings) instead of the default random seed, you can reuse that same seed again the second time to lock in the same style of output
Hmm. I'm pretty impressed at the decent lyrical performances. It's understandable and pretty decent in terms of the actual words. The actual sounds of the instruments are still very wonky, though....
Hmm. I'm pretty impressed at the decent lyrical performances. It's understandable and pretty decent in terms of the actual words. The actual sounds of the instruments are still very wonky, though.
I asked it to do a song in the style of Yuki Kajiura, which I think is a pretty good test since she has a particularly unique style that's hard to pigeonhole. The first song was a weird cinematic opera - a definite fail. But the second one was fairly close, in that it had the electronic backing and female vocals.
How did you request a song in the style of Yuki Kajiura? I asked for music in the style of Chopin, but it said it couldn't use artist likenesses without permission, automatically changed my prompt...
How did you request a song in the style of Yuki Kajiura? I asked for music in the style of Chopin, but it said it couldn't use artist likenesses without permission, automatically changed my prompt to "western classical," and charged me two credits.
I would imagine it is simply because they don’t know who she is. But I used “in the style of Yuki Kajiura” which might have helped. Here’s what I came up with:...
I would imagine it is simply because they don’t know who she is. But I used “in the style of Yuki Kajiura” which might have helped.
I've been having a lot of fun with Suno and at first blush this seems like it's a lot higher-quality and more powerful in terms of the control it gives you. Suno definitely has some quality...
I've been having a lot of fun with Suno and at first blush this seems like it's a lot higher-quality and more powerful in terms of the control it gives you. Suno definitely has some quality issues. My sole gripe is that Udio is way slower to generate, but if the tradeoff is actually quality output then that's probably not too bad of a drawback.
One of the settings under the advanced features menu is a generation quality slider. It defaults to High, with the four available options being Fastest, Fast, High, or Ultra. So you can always try...
One of the settings under the advanced features menu is a generation quality slider. It defaults to High, with the four available options being Fastest, Fast, High, or Ultra. So you can always try turning that down if you want to experiment more quickly!
I'd be tempted if there was a way to land on a particular sound and style and then use it as a reference for a higher-quality piece, but as it is I think it's better to eat the wait time and land...
I'd be tempted if there was a way to land on a particular sound and style and then use it as a reference for a higher-quality piece, but as it is I think it's better to eat the wait time and land on something more usable. Higher quality means, generally, fewer iterations to land on something that sounds decent. Even sticking with high it only took me about 20-25 generations to make a decent 2 minute song.
I haven't tried this yet myself, so take this with a grain of salt, but that should be possible if you set a custom seed for the fast gens and then reuse the same custom seed+prompt at higher...
I'd be tempted if there was a way to land on a particular sound and style and then use it as a reference for a higher-quality piece
I haven't tried this yet myself, so take this with a grain of salt, but that should be possible if you set a custom seed for the fast gens and then reuse the same custom seed+prompt at higher quality later on. (You can specify the seed under the advanced settings menu, which defaults to -1 aka random)
I think it'd be more consistent to generate the intro, then use that to generate a higher quality verse/chorus before cropping and regenerating the intro? It just seems like a pain in the butt, haha.
I think it'd be more consistent to generate the intro, then use that to generate a higher quality verse/chorus before cropping and regenerating the intro? It just seems like a pain in the butt, haha.
Also wow. The lyrics can get a bit simple/campy, but I'm impressed with how sophisticated the vocals are. It made a ballad for my dog I recently lost and it hits in the feels:...
Also wow. The lyrics can get a bit simple/campy, but I'm impressed with how sophisticated the vocals are. It made a ballad for my dog I recently lost and it hits in the feels: https://www.udio.com/songs/tRc5KepKRPwGFc2jzpq3Yr
I love this so much! I was feeling festive and made a holiday classic (with a twist). The quality of the output is consistently blowing me away, it feels magical.
I love this so much! I was feeling festive and made a holiday classic (with a twist). The quality of the output is consistently blowing me away, it feels magical.
This is much more to my liking than Suno. Suno had some issues with output that drove me nuts - weird machine howly noises, instruments blending together into an airy, piercing sound, among others...
This is much more to my liking than Suno. Suno had some issues with output that drove me nuts - weird machine howly noises, instruments blending together into an airy, piercing sound, among others - and its methods of getting some consistency didn't feel like they worked very well. Udio doesn't seem to be as straightforward combining genres in the prompt, but the other settings/tools absolutely make up for that + the results have more interesting choices from what few I've generated thus far. Time to click all the buttons, I'm excited to mess with this.
Edit: I'm dumb and didn't see Manual Mode, and wow does this sound better
Again edit: Here are a couple of examples of what I could get it to do. You can do a bunch of the same kinds of prompting in the lyric box, but similar to Suno that doesn't seem to be as consistent as I'd hoped. Still, cool shit, I'm enjoying myself
I run an F1 pool and I thought it'd be fun to have a motivational theme for it... [link] not too bad. edit: here's one for Google Sheets! https://www.udio.com/songs/qh3Sutvm5hJcj42xLJy6n1
I run an F1 pool and I thought it'd be fun to have a motivational theme for it...
Wheels on fire, feel the speed
Tearing through the fight's intensity
Victory's in the roar, oh the lead
Crossing that line, destiny
[Chorus]
Champion's rise, taste the win
Never back down, let the race begin
Engine's scream, adrenaline's surge
In the zone, on the edge like a purge
Lights flash by, tunnel vision clear
Overcome the fear, victory's near
[Verse 2]
Heart's pounding, push the limit
No regrets, just the race spirit
Track's alive, the crowd's wild
You're a legend, can't be denied
Sweat and steel, it's all you need
Born to race, destined to lead
[Chorus]
Champion's rise, taste the win
Never back down, let the race begin
Engine roars, feel that power
Every second, time's your scepter
In the heat, true grit won’t cower
Push the pedal, racing faster
[Bridge]
Hit the curve at breakneck speed
Turbocharged, you take the lead
This is the first such generator I've seen that not only makes coherent, realistic-sounding music -- Jukebox was awful in comparison, and even Suno has a weird synthy quality to the vocals -- but also makes songs that are genuinely catchy.
I did a post on it for MetaFilter a few weeks ago; here are some highlights that I came across:
Dune the Broadway Musical [Showtunes] - Baby On Board [Barbershop] - Carolina-O [Indie Country] - Sabrosito Amor [Latin] - Rising Sun Gospel [Soul] - Allegro Consort in C [Classical] - You Spilt a Coffee on my Dog [R&B] - Potion Seller [60s Folk] - I'm Not Your Star [Screamo] - SNES Greensleeves [Chiptune] - Syncopated Rhythms [Jazz] - Tavern Serenades [Fiddle] - My Tamagotchi died in '98 [Country Pop] - Senna Tea Blues [Bluegrass] - Unexpected Item in Bagging Area (A Cowboy's Lament) [Americana] - Herb's Whisper [Hip-hop] - Metropolis Pt. 3 [Prog metal] - F**k You Elmo [Acoustic Guitar] - Lorem Ipsum Dolor Sit Amet [Orchestral] - ムーンライト【.】【3】【1】[Vaporwave] - Dreaming Miku [Vocaloid] - The Deku Tree’s Decree [Broadway] - Website on the Internet [50s A Capella]
I still find myself humming "Website on the Internet", even weeks later.
Jesus. As a musician who wrote a few years ago about AI's ability (or lack thereof) to make art, this caught me wildly off-guard.
What was your prompt for "Website On the Internet"?woo reading comprehensionI didn't make any of them, just pulled my favorites from the top charts, but the tags for that song list: Barbershop, Rhythmic, Happy, Melodic, Playful, Vocal group, Warm, 1950s Barbershop, Rhythmic, Happy, Melo...
I was playing with Suno to make a version of the New Zealand national anthem.
I asked for polka as a joke, but it made something actually listenable, rather than a chore to suffer through prior to sports games, school assemblies, or state events.
The Māori words are butchered, but in the same way that most people would sing them.
The English verses (or at least the more well known second verse) begin with the same loudness and chorus that happens as people join in singing since that's the words they know.
“You Spilt a Coffee on My Dog” is a standout for me. It sounds like a Robyn song.
I genuinely adore the lines “ain’t no mistake you dropped that brew / what my dog ever do to you?”.
A lot of AI-generated content has a sort of soulless uncanny-valley effect on me that’s off-putting, but this is great. I’m legitimately into it.
Um... wow.
I've known about the advances in generative AI of course, but 'being aware' is a different experience from listening to music that has vocals I can't discern being computer generated while at the same time the music is actually good to my tastes.
Well. It's about time I personally got something nice from Skynet before... you know... Skynet does Skynet things.
I… Uhhh… wow
My honest reaction
I even asked it to sing a song about a city in Portugal, and I wrote my prompt in English, I didn’t expect it to actually deliver a portuguese song using Fado
This is awesome! The editing tools are really neat - you generate your song in increments of ~30s or less -- well, technically 30s each time, but you can trim out any parts of those 30s you don't want to keep -- and best of all imo, you can redo your prompt at each stage, to evolve the style of the song as you go while trusting that it'll still make the whole thing fit together in the end.
Here's my first take at a full-length track: an overdramatic symphonic metal/gothic rock song about being sad that your cat won't come over to cuddle you lol (I had fun with it)
Do you get any fine-tuning editing of the lyrics?
There are a few options for that:
Hmm. I'm pretty impressed at the decent lyrical performances. It's understandable and pretty decent in terms of the actual words. The actual sounds of the instruments are still very wonky, though.
I asked it to do a song in the style of Yuki Kajiura, which I think is a pretty good test since she has a particularly unique style that's hard to pigeonhole. The first song was a weird cinematic opera - a definite fail. But the second one was fairly close, in that it had the electronic backing and female vocals.
How did you request a song in the style of Yuki Kajiura? I asked for music in the style of Chopin, but it said it couldn't use artist likenesses without permission, automatically changed my prompt to "western classical," and charged me two credits.
I would imagine it is simply because they don’t know who she is. But I used “in the style of Yuki Kajiura” which might have helped.
Here’s what I came up with: https://www.udio.com/songs/v8tkiHyhoeKi31yo4GFdCx
I tried to have it make an intro for it but both of them started with rap, which was kind of weird.
I've been having a lot of fun with Suno and at first blush this seems like it's a lot higher-quality and more powerful in terms of the control it gives you. Suno definitely has some quality issues. My sole gripe is that Udio is way slower to generate, but if the tradeoff is actually quality output then that's probably not too bad of a drawback.
One of the settings under the advanced features menu is a generation quality slider. It defaults to High, with the four available options being Fastest, Fast, High, or Ultra. So you can always try turning that down if you want to experiment more quickly!
I'd be tempted if there was a way to land on a particular sound and style and then use it as a reference for a higher-quality piece, but as it is I think it's better to eat the wait time and land on something more usable. Higher quality means, generally, fewer iterations to land on something that sounds decent. Even sticking with high it only took me about 20-25 generations to make a decent 2 minute song.
I haven't tried this yet myself, so take this with a grain of salt, but that should be possible if you set a custom seed for the fast gens and then reuse the same custom seed+prompt at higher quality later on. (You can specify the seed under the advanced settings menu, which defaults to -1 aka random)
I think it'd be more consistent to generate the intro, then use that to generate a higher quality verse/chorus before cropping and regenerating the intro? It just seems like a pain in the butt, haha.
Also wow. The lyrics can get a bit simple/campy, but I'm impressed with how sophisticated the vocals are. It made a ballad for my dog I recently lost and it hits in the feels: https://www.udio.com/songs/tRc5KepKRPwGFc2jzpq3Yr
I love this so much! I was feeling festive and made a holiday classic (with a twist). The quality of the output is consistently blowing me away, it feels magical.
Hopefully we get open-source tools that can be integrated into a DAW workflow at some point.
This is much more to my liking than Suno. Suno had some issues with output that drove me nuts - weird machine howly noises, instruments blending together into an airy, piercing sound, among others - and its methods of getting some consistency didn't feel like they worked very well. Udio doesn't seem to be as straightforward combining genres in the prompt, but the other settings/tools absolutely make up for that + the results have more interesting choices from what few I've generated thus far. Time to click all the buttons, I'm excited to mess with this.
Edit: I'm dumb and didn't see Manual Mode, and wow does this sound better
Again edit: Here are a couple of examples of what I could get it to do. You can do a bunch of the same kinds of prompting in the lyric box, but similar to Suno that doesn't seem to be as consistent as I'd hoped. Still, cool shit, I'm enjoying myself
I run an F1 pool and I thought it'd be fun to have a motivational theme for it...
[link]
not too bad.
edit: here's one for Google Sheets! https://www.udio.com/songs/qh3Sutvm5hJcj42xLJy6n1