Menu
What's new
By:
Filters

Made with AI: Suno, Udio and others

The more I listen to Udio stuff the more both the strengths and weaknesses stand out. The timbre and subtleties in some moments of individual performances are just stunning: world class and beyond. I tried prompting for an imaginary gospel-funk version of one of my songs and the disjointed mess that came out had one of the most glorious, crunching grooves I’ve ever heard: just perfection. Another song I clicked on randomly had one of the best Zeppelin/Sabbath rock riffs I’ve heard in ages: all of it from the tones to the playing to the note choices to the slurs was of classic standard.

When it comes to the purely aural subtleties of groove, mix and tone, I don’t think even the very best are going to be able to compete with the levels reached by AI. You’ll either use it as part of the process or you just won’t hit the same levels. It reminds me of the heightened mixes some movies have in their soundtracks, like Guardians of the Galaxy, but more so. It also has something in common with sampling: perfect little gems of sound-in-a-moment that can only be found and not made and also have that heightened, unreal, highly stimulating quality.

It’s going to be interesting to see how listeners take to it once it’s mature enough to make cohesive pieces that really hold together (likely with human intervention beyond just prompting and tweaking parameters). If it becomes widespread in media, listeners may come to expect those levels of hyper stimulation.

The weaknesses, I’m hoping they stay there for a while. The drunken, rambling quality, the complete tin ear for melody, melodic development and lyrical phrasing.
 
I happened to be cooking just now, so I prompted: "a choir singing about cooking"

And it provided me with two tracks, one which is randomly something like a 80s prog metal band like Dio -

www.udio.com

Udio | Choral Feast by jonathanwing

Make your music
www.udio.com www.udio.com

I find that to be shockingly close to any of those 80s pop metal songs, and while it's kind of funny, "Spices of the earth, now rise" feels quite idiomatic to that anthemic style.

And the second which I was anticipating moreso, a traditional church choir -

www.udio.com

Udio | Choir of Flavors by jonathanwing

Make your music
www.udio.com www.udio.com

Interesting here is if you look at the lyrics, it seems like they only intelligibly sing a couple of the words. The rest sounds like gibberish. Maybe this particular model had been trained on various traditional choirs using languages from Latin to German and so it just makes approximate sounds?

I agree with others though, this technology would be best harnessed by composers for DAWs and Notation software to make better sounding mockups more quickly, especially for stuff like choirs and soloist singers where as someone else said we are relegated to ooh's and ahh's for the most part. As well as the ability to pull off more convicing aleatoric and glissando passages.

I wanted to see if I could feed it a prompt for a soloist with specific key & tempo instruction - "a capella singer in the style of pop, in the key of Cm 90bpm" - just to see if one could generate something for use in a song, merely out of technical curiosity. This absolutely did not work. It just created some pretty horrible pop tracks with a full band, at various tempos and different keys. I think it took what I said about "key of Cm" and wrote extremely awkward lyrics literally about melody and harmony.

Other than kitsch factor, I don't really see the point of technologies like this unless designed for composers/producers as a tool to aid in production, rather than replace the artist entirely. The world already has enough shitty music, so there's absolutely no need and purpose to continue the flood of shitty soundalike music with pointless comedy lyrics.
 
PedroPH said:
I didn't say that. Of course that's not the case. But there is a lot of music made by professionals that is just like what I described, and that kind of music is much more of a candidate to be generated with AI than the compositions of "any hobby composer who composes classical/new-age/film-music with only chord progressions and a melody". In addition to this, an amateur composer can't possibly have his career as a composer threatened by AI because he doesn't have such a career, obviously. So, why mention, hobby composers in this context?
Click to expand...
You said note clusters before: I don’t think note clusters or any other fancy technique are going to save anyone. All it takes is training a quick LoRA on RunPod for a couple of bucks in GPU time (low-rank adaption) with a few examples to teach a model how to ape an artists style. More than anything it’s the name of the artist and reputation that protects them: they’re still the only one that makes the genuine article.
 
www.udio.com

Udio | Tears of Clementine by Yomi

Make your music
www.udio.com www.udio.com

Here's an attempt from last night. The initial freewheeling result then took up to 80 generations to expand into this piece. Rather than fight the model's drunken verbosity, I found that a genre like Prog Rock allows for it to be made a feature rather than a bug.

It still required a couple of hours of this workflow:

1. Fiddle with lyrics and [captions] in hope of influencing the result
2. Pull arm of one-armed bandit to generate new section
3. "Not my tempo!"
4. Goto 1

Despite enjoying the final result, I don't recommend the experience of extending a piece due to the lack of steerability: every generation has flaws and this tends to include the best and most interesting ones.
 
I think we should signup for all the AI music related chatrooms/social media/forums etc and and pretend that we dont know anything about music and say we have been uploading tons of ai music to YouTube/spotify/ditrokid royalty free sites and start saying that "after 400 tracks uploaded ive made about $100 in 6 months and that its taken around 1000 hours to do all that. "
Or some variation on this.

Because the way I see it, these people who are fascinated with ai music just want to do this. Use the lowest common denominator way to make some money and have no idea how the music industry works. Lets say I upload ai generated music that sounds exactly like Taylor swifts new album... not the same song but ai generated music got all the "right" stuff to make music that sounds like if she made it and sounds very good.

What then?

yep... ai enthusiasts who are the same as those NFT guys peddling bored apes jpegs are the ones who are creating this demand for prompt to amazing music apps with the intention to sell the music and get money. Just like they said about using crypto to "fix" the paper work involved in health care, or create secure voting systems and so on...basically a complete misunderstanding of how society works and just ASSume that if NFT and crypto is complicated then "those" other things should be easy to "fix".

Case in point; the music business relies a lot on marketing and the artists itself. someone pushing social media buzz and concerts and so on to get a little ahead a little at a time.
 
wing said:
Other than kitsch factor, I don't really see the point of technologies like this unless designed for composers/producers as a tool to aid in production
Click to expand...
To be the devil's advocate for a minute: suppose your favorite Abba album is "Arrival", which you find rather special relative to the rest of their entire catalog. If so, you might enjoy a 3-hour AI playlist of "Arrival" soundalikes while you're cooking or cleaning. Maybe one day AI will be able to do that.

I think anyone who uses music for background while focusing on something else might be interested in some playlist like that.
 
youngpokie said:
To be the devil's advocate for a minute: suppose your favorite Abba album is "Arrival", which you find rather special relative to the rest of their entire catalog. If so, you might enjoy a 3-hour AI playlist of "Arrival" soundalikes while you're cooking or cleaning. Maybe one day AI will be able to do that.

I think anyone who uses music for background while focusing on something else might be interested in some playlist like that.
Click to expand...
I get your point, but the scenario sounds quite strange, especially considering that there is no lack of music in the world to listen to. A lot of streaming platforms have pretty good algorithms for playlists and recommendations in a similar style, so I don't think there are many people out there thinking "I really wish I could listen to a reprisal of this song done hundreds of different ways for the next 3 hours." I think people are more interested in finding new recommendations based on that song, which Spotify, Pandora etc already do.
 
wing said:
so I don't think there are many people out there thinking "I really wish I could listen to a reprisal of this song done hundreds of different ways for the next 3 hours."
Click to expand...
Well, it seems to be a big enough thing that journalists write about it...

1713561223406.png

I never listen on repeat, but my Spotify recommendations always suck. I like a certain type of sound and vibe, but Spotify always assumes I like the music from the decade of the song I put a like on.
 
youngpokie said:
Well, it seems to be a big enough thing that journalists write about it...

1713561223406.png

I never listen on repeat, but my Spotify recommendations always suck. I like a certain type of sound and vibe, but Spotify always assumes I like the music from the decade of the song I put a like on.
Click to expand...
Yeah, I listen to songs I like on repeat from time to time, but I think AI generated remix/reprisals of it over and over would drive me nuts. But sure, maybe someone out there would love that, to each their own. Still seems like an odd justification for it given that we're not running out of anything to listen to.
 


I love this, some of the lyrics are so enticing…. Yet at the same time I felt revulsion that I was being duped by a machine.

Time to read some philosophy.. my mind is blown.
 
I wrote some lyrics (only spent 15 minutes on them and English is not my native tongue..)
I did not pay attention to the syllable count and it's interesting how it handled it on the 3rd line.

I hate that I love the song even if I still couldn't get it to follow a verse-chorus structure

www.udio.com

Udio | Urban Destiny by sfpx | AI Music Generator - Official Website

Listen to Urban Destiny by sfpx on Udio. Discover, create, and share music with the world. Use the latest technology to create AI music in seconds.
www.udio.com www.udio.com

There were a couple of different paths to the song that were better but I rejected them because Udio was adding some random gibberish. There is still some gibberish in the song I published but since it was toward the end I kept it.
 
Last edited:
So far the Udio / Suno experience for me was mostly like:

Oh wow that part sounds great, scary what AI can do. I feel kinda outperformed. Am I useless now? But I would do that bit a little different, also the mix and audio quality sucks.

So it serves as inspiration.
 
Uncle Peter said:


I love this, some of the lyrics are so enticing…. Yet at the same time I felt revulsion that I was being duped by a machine.

Time to read some philosophy.. my mind is blown.
Click to expand...

This account was banned by YouTube, I heard reports of people losing their account, artists are denouncing Suno and Udio songs for similarities, sometimes for plagiarism, it seems like it will be difficult for this to prosper.
 
You must log in or register to reply here.
Top Bottom