All depends on the genre you want the pads to be in. But here is a minimal UK garage song with said juno style pads

https://suno.com/s/YLofJQ6gZmObjEc7

 

if were already f*cked why dont we learn how to use the tools to further our creativity in these times? Vibe code a plugin, generate a stem, game plan a rollout, generate cover art. etc etc..

We can shake our canes at AI all day. Or we can learn how to use it and capitalize on this moment to free our creative ideas to see the light of day!

 

Capitalism and the so called "communist" countries are just different sides of the same coin when productivism, exploitation, alienation, technology, war and growth are controlled by and used on behalf of the power elites (economic, military, technocratic).

The point is that if these elites are in oligopolistic control of technology, growth, etc. they will try to use all this to increase and mantain their power (political and economic) - capitalism is just a way to achieve this in the West (with the facade of a "free" market).

 

Last edited: Aug 26, 2025 at 12:57 AM

But these AIs don't create something out of nothing... they recombine and vary existing musical and textual patterns that have been "digested" and stored through statistical learning. The real feat lies in the neural network's ability to understand the complex relationships between these elements, combining them in a way that sounds coherent and, sometimes, surprising (thanks to the insertion of randomness and variations!), with audio sample results being generated quickly with increasingly remarkable quality! So...

 

Last edited: Aug 26, 2025 at 12:53 AM

Kind of. No audio is stored or used in the process. Its just like GPT it predicts the next token. So it takes in millions of songs and gets a understanding of them based on a tagging system. Then the algo uses a noise profile to start from. Think of it as a block of clay of white noise.

it shapes the noise to be one millisecond of audio that is based on the context you give it. Then it continues that process for the duration of the song.

Kind of how a series of still photos together become a movie scene.

No Audio is being used in generation. Just the overall concepts of the audio. Its just white noise being guided by machine learning to generate audio out of noise based on its understanding of music guided by a prompt or another piece of audio.

 

I see, the AI doesn't store the patterns themselves, but rather the relationships and statistical probabilities between all musical elements, in order to perform an intelligent and statistically informed recombination of learned patterns...

Got it. So AIs are basically masters of imitation and recombination, right? They're still just constructs with no intention, no emotion, no life experience, no culture, and no ability to make unexpected connections between unrelated concepts, like a human would. And this is all just for songs, right? Honestly, I'll be really impressed when an AI creates something on the level of Stravinsky, Villa-Lobos, Thomas Adès, or John Adams... And believe me, I'm not being ironic, 'cos I truly think this will be possible one day! but probably not in my lifetime... or yours... or your kids'... but maybe your grandkids'!

 

Last edited: Aug 26, 2025 at 1:43 AM

"It's so amazing when people tell me that … electronic music has not got soul. And they blame the computers. They got the finger pointed at the computers like, "There's no soul here." … You can't blame the computer. If there's not soul in the music, it's because nobody put it there. And it's not the tool's fault."

I hate spitted "AI music" as much as the next neighbor but c'mon. The diffusion models realm It's a new media in it's infancy. Is like seeing the first lumiere brothers movie and say: fuck it, I'm going back to the theather were the true art resides with real people and not these ghost trains splashed over a white blanket.

Don't be ludite, just because any moron can take a camera and make a video it doesn't invalidate the audiovisual media as a whole.

 

More that deep "understanding", is a deep learning of methods of diffusion from noise to signal. Is not AI at all, we have been fooled by that.
Just take a song, tag it, then add white noise in little intervals until all you have is white noise. Do the same thing with millions of songs. Then teach the model how to take that white noise, and based on tags (prompts from the user) diffuse it into something that has similarities with all or some of the songs with that tagging. The model doesn´t understand shit. It just diffuses different sets of noises (seeds) to signal (results).

 

Yes, AIs don't have consciousness or intuition... but for specialists and computer scientists, diffusion models are indeed a type of AI, and specifically, a subfield of machine learning. In short, the creation would, in reality, be an extremely sophisticated algorithm, trained to follow mathematical rules. It generates something that sounds like music because its database taught it to recognize patterns and associations ('noise' phase vs. 'diffusion' phase), not because it 'understands' what it's doing! Impressive!

 

I don't know why people just care so much about Suno trash songs. always they refer to Suno when It comes to music despite there are better alternatives. Is it just because It is popular?
Producer aka Riffusion can already make catchy songs yet still consistent than Suno trash quality
Sonauto with dedication can do it. Udio is superb, but It needs only better mastering in the transients and latency.
People are just taken by one AI platform, makes me wonder how aware the new generation.
Seems like they are noob
and this came from a very popular audio forum
I don't think AI will kill creativity, but it will challenge creativity
It will raise the roof. Creativity has no limits unless you submit to AI defaults, then you are OWNED.
Now the song is trending because the musician felt impressed like a noob without any good critic.
The reverb is good but so overpowered if this is not remastered before uploading
the vocals stop on long verses in an unexpected way though it stretches itself but not good melisma at all.
Chorus is garbage
no good variety at all, neither good transition
sounds like a demo ballad than a pop song
only the piano sounds a little clear
no space from some vocals, despite it requires.
Suno is trash at freestyle lyrics
also very generic female pop vocals, not interesting at all.
So many losers to AI deepfakes. Musicians turned out to be very loose
to call this is good out of the box is haunting me.

 

this is an example of image diffusion its the same theory with audio diffusion. Take image (or audio) encoded with a pile of text that describes it add noise the repeat until the model is trained. Now enter text you want the process goes in reverse it will decode from the noise and give you what you want based on your text prompt. You can add audio to prompt too!

Thats the thing I think most are confused with when using these AI tools.

They think its "Make me song that sound good"


But you can put all of those things listed into a song with or without AI. The cool thing with the AI tools lets say youre a man who needs soulful female backup vocals on your song. You can "Prompt" with your original vocal take and imagine that vocal into any genre from any generation as a female backup vocalist. Apply that to anything you can imagine! You now have the ability to use a tool like a amp simulator to take source material and take reimagine that source material into anything! To me thats a powerful use of the new tech.

 

Last edited: Aug 26, 2025 at 5:10 AM

This debate is like "why do people use studio one, FL studio, ableton, logic.... debate. They all have their strengths and weaknesses.

I have been using them all plus some others. They all give their own flavor of results. The power is combining them!

Sonauto generates one good song every 20 but their lyrics engine and sound quality also certain genres sound better than others.

Sometimes i take a suno gen and run it through riffusion just to upscale the sound.

Who cares what people use as long as its what they want in the end?