AI Generated / Assisted Music

Tried playing with Udio again and it really does feel like a slot machine, you click the button and hope for the best even if you upload a sample for it to start with. This song somehow ended up with a rock vibe to it at the end, even though I didn't prompt for it. It still kind of works (mostly), but it's not 100% what I was going for.

 
I think I've finally cracked the mystery of Udio. Seems that it works best whenever you actually start off with a simple prompt, pick what you like, and then change your prompt for each section, to incorporate whatever it is that you want to hear in terms of styles or vocals. Also, a nice trick is to start something in Suno, export that, export it to Udio and finish it there, so then you get the best of both worlds.
 
It was bound to happen.
Yes, because people are greedy. If a human "trains" on music and then creates music that is similar to those artists then it's fine, they can even teach other people how to do it too, but heaven forbid that AI does it.
 
Yes, because people are greedy. If a human "trains" on music and then creates music that is similar to those artists then it's fine, they can even teach other people how to do it too, but heaven forbid that AI does it.
Because money can be made by labels and agents through people. They can not make money off of user generated content using AI.
 
At the top of the group of things that I hold in no high regard would be record labels. They are as criminal as can be screwing over artists left and right.
 
Yes, because people are greedy. If a human "trains" on music and then creates music that is similar to those artists then it's fine, they can even teach other people how to do it too, but heaven forbid that AI does it.
Ye, I'm going to hard disagree with you here.

You cannot really compare machine learning tools to a human. It takes talent and lots of time for a human to "train" whereas these AI tools are just going to mass manufacture music to such a degree that it becomes meaningless and floods the market. While all the way these companies are going to milk it for as much as they can.

I'm still super impressed with what, specifically Suno, has achieved but it's clear their training dataset was massive and certainly used copyrighted works.
 
Ye, I'm going to hard disagree with you here.

You cannot really compare machine learning tools to a human. It takes talent and lots of time for a human to "train" whereas these AI tools are just going to mass manufacture music to such a degree that it becomes meaningless and floods the market. While all the way these companies are going to milk it for as much as they can.

I'm still super impressed with what, specifically Suno, has achieved but it's clear their training dataset was massive and certainly used copyrighted works.
Ok, let's use another analogy. This is like Kodac suing Minolta because a photocopy machine has the potential to copy a copyrighted photo and "flood the market" with lower quality content.
 

I feel using Auto-Tune is the same as using AI. Yes I know everyone is going to tell me I am wrong. In the article she does say that it made her lazy. You can hear in many of the AI Generated songs it sounds like an old version of Auto-Tune is being used.

With Auto-Tune anybody, even those who can not carry a tune to save their lives, can sound like a pro.
 
I see audio input is now also available to free users... Sounds like fun.
 
I see audio input is now also available to free users... Sounds like fun.
It can be, just temper your expectations. From how it was promoted it sounded like the next best thing since sliced cheese but in reality it's just "okay". There are times when I can't hear anything that even resembles the input audio. Like with this Running track I just shared, I had to do 4-5 versions of before I got a version where you can even hear the rhythm of my footsteps in the actual song.

Also, it seems you need very high end recordings, with little to no background noise or the AI will get horribly confused and it will sound like you opened a portal to hell.
 
It can be, just temper your expectations. From how it was promoted it sounded like the next best thing since sliced cheese but in reality it's just "okay". There are times when I can't hear anything that even resembles the input audio. Like with this Running track I just shared, I had to do 4-5 versions of before I got a version where you can even hear the rhythm of my footsteps in the actual song.

Also, it seems you need very high end recordings, with little to no background noise or the AI will get horribly confused and it will sound like you opened a portal to hell.
You sounded like a heavy smoker at the start there :ROFL:

Song is good though.

Hate the start of this one though:

I really get annoyed by the ice cream vans track. Drives my dogs crazy as well. Gets them to howl like they are in a lot of pain.
 
Top
Sign up to the MyBroadband newsletter
X