Ahh well, I finally got tired.
If rebellion 40 years ago was using electric guitars, it's probably using AI now.
With all the AI-driven music, I've been trying to make it myself, and as a computer scientist, music enthusiast, and a bit of a rebel, I wanted to truly test what AI can do, how far it can go—it's just an experiment.

 

Last edited: Oct 18, 2024

Pretty wild... pure 80's vibe, like Duran Duran and INXS had a baby.

I like it even though a part of me still struggles with the idea of liking something that isnt technically made by a person or people I cant go see at a show. I know that can be said about a lot of music but when lyrics are being sung, it hits different.

 

Last edited: Oct 19, 2024

It´s a G ..and then F. ..and then it´s a G... and then a F . and now comes a G ! ... and once again an F

But when u prompt it further , I bet it can make more.... and that scares me.

 

... what program did you use for the vocals? I should mess with those for my stuff, ive been making pop music with those acapella packs.

 

It is 100% SUNO.
1) As a generative AI, it starts with a pseudo Random seed (but in SUNO you can extend(an existent one(upload a melody with G ..and then F), which it is a seed itself))
2) Thumbs Up Down guides the next generation, penalizing bad generations.
3) prompt changing styles
4) Re-seed
For example :
from a real seed(composed with Ableton)

to a AI version

 

the real problem is the terms of use and you dont own your song

 

Not really.
Suno Terms says" if you are a user who has subscribed to the paid tier of the Service, Suno hereby assigns to you all of its right, title and interest in and to any Output"
https://suno.com/terms

And before you start talking about laws and posting links to the Sony vs. SUNO lawsuit, etc., let's better focus on the musical aspect, shall we?

 

Cool! I'm at work right now, ugh, but I'll check Suno out sometime when I'm bored. Thanks!

 

the thing is very interesting and the results are pleasant well done!

i have "played a bit" with the free account. What I would like to know is if the AI can be guided with some "true musical tools" lol I mean does the AI read a simple score with melody for vocals for example? Or a chord progression on a double piano staff?

I would like to give the AI indications in "music form" not in words. Is it possible?

 

I wasted the last year allowing AI to think for me. It was great!! It did things I wouldn't have ever thought of. But there was one vital ingredient missing. Me!! I'd forgotten the joy of creating a song, Maybe AI can write a better song but I missed being part of the process.
Because if you're a musician, write your music and make it your own.

 

I agree.
I just want the AI to help me, not to replace me.
I am trying to find the latter.

 

yes I bet there will be a formal structure to prompting were there will be certain non-ambigous grammar and syntax to be incredibly accurate when giving instructions, we can parse this language and feed it to the AI in a very optimized form that the AI can understand clearly, now writing stuff is annoying I bet a genius engineer will figure a visual way to combine ideas and blocks into a coherent system where you get the correct level of abstraction and not get bogged down with details but under the hood this visualization is just translated to the language, in this visual system you can choose your own sounds and timing , you even have control over the waveforms you can literally tell the AI to remove higher frequencies and normalize the volume pretty amazing stuff, now you will get a few talented individuals that can use this visual style to create really good music , they dont prompt directly per say because its too hard they move blocks on the screen to customize their required sound

 

This is what I was talking about, profesional musicians , producers , early adopters ,using AI

 

I don't see a problem using AI, as long as you don't lie and say it's your work, but the AI did everything. I write lyrics. Work out the chords for the melody and then feed it to the AI (and it takes many renders to get what you want) and then if it generates something that I can use as a reference, then I use it. I will never post something and say it's all my work if I, for instance let the AI do the music and the voice. I will specify that I did the lyrics and the melody and that the AI did the music and the vocal. That is my song. I wrote it. Doesn't matter if the AI did the music and the vocals. It's my song and I own it and register it as my own.

 

With good prompts and feeding with audio i can generate better melodic samples than Splice on Suno. I mean, at least the result is not poorly sound designed as some bad ones from Splice or other sample packs from untrusted devs. And im always using my own work to feed the AI, chords, instruments, everything, so i can get some extra stuff to finish my cake.

I don't think a work 100% arranged and sang by IA sounds good at all. But you can use it plenty of other ways, even turning some stuff that sounds a little wack to MIDI and using UVI Falcon.

 

reminds me of those "good old 80's", yes. But it's way better than what ever Ms. Swift does. Matter of taste of anyway ...

 

After a half-year with Suno, I can recognize their patterns everywhere, as well as the junk noise that appears in every song. And the voices sound like some corporate guy or girl singing at a coffee shop in the same room with the same reverb all the time. If the corporate person is younger, it sounds like someone is suffocating him or her to death.

 

I read the terms, word by word
thats why I got out weeks ago
I'm not trying to start a war here, but its important to know what are the limits of the tool, because of later frustration