AI Generated / Assisted Music

Well, Suno just released their "Audio to Audio" feature... can't wait to try it out! Basically, you can take any sound and turn it into music.
 
Sounds interesting... Will check it out when I can.
So far, it's not quite what I expected. It's pretty much the same as Udio's (also new) audio extend feature, where it will continue a sound from a sample you upload. It will try to apply the style you give it in the prompt, but from my early experiments, it either sort of loops the sample, or creates music that sounds nothing like it.

Edit - found it works much better with clean samples, any kind of hissing or background noise in the clip or video will confuse the AI horribly. I think I've heard voices from Hell tonight, just from trying to use an unclean sample with Suno...
 
Last edited:
In its current incarnation, this feature looks like it's made to have people burn through their credits. It rarely does what you prompt it to do, the below "test" is me humming a well-known theme song and tried to turn it into a rock / metal tune. The first several tries ended up just looping and remixing my vocal and eventually I got some kind of guitar riff, but it was more acoustic-bluesy and not the style and genre I asked for. Suno also liked to generate random lyrics and vocals despite me specifically setting it as an instrumental track. The initial pass of the "hum-to guitar" also included my voice, so I downloaded that song, snipped off 15 seconds of guitar only, and fed that through Suno again. I tried once more to make it into a metal / hard rock song but Suno kept it to the acoustic-bluesy style despite my tried. Anyway, the result is alright - it's just not what I asked for.


Udio's feature is really for "finishing" songs, unlike Suno, it doesn't even try to convert your input into something else, it just extends it and adds some layers and extras as the song progresses, but it keeps whatever the original sample was, so if you hum something, the entire song will include samples and remixes of your humming.
 
In its current incarnation, this feature looks like it's made to have people burn through their credits. It rarely does what you prompt it to do, the below "test" is me humming a well-known theme song and tried to turn it into a rock / metal tune. The first several tries ended up just looping and remixing my vocal and eventually I got some kind of guitar riff, but it was more acoustic-bluesy and not the style and genre I asked for. Suno also liked to generate random lyrics and vocals despite me specifically setting it as an instrumental track. The initial pass of the "hum-to guitar" also included my voice, so I downloaded that song, snipped off 15 seconds of guitar only, and fed that through Suno again. I tried once more to make it into a metal / hard rock song but Suno kept it to the acoustic-bluesy style despite my tried. Anyway, the result is alright - it's just not what I asked for.


Udio's feature is really for "finishing" songs, unlike Suno, it doesn't even try to convert your input into something else, it just extends it and adds some layers and extras as the song progresses, but it keeps whatever the original sample was, so if you hum something, the entire song will include samples and remixes of your humming.
I heard a faint bit of Mission Impossible but the rest sounded fine to me.
 
I heard a faint bit of Mission Impossible but the rest sounded fine to me.
Well, that's the tune I hummed for the input. It was supposed to be hard rock and not blues. Also had to do many generations to get one that doesn't have vocals or my original voice sample in it.
 
feature looks like it's made to have people burn through their credits
Most of the time it does that I have found. It just messes up the custom lyrics and then cuts off on the outro. If you rerun the same prompt, it comes out fine. It may sound like a conspiracy theory, but it happens too often to be a coincidence.
 
Most of the time it does that I have found. It just messes up the custom lyrics and then cuts off on the outro. If you rerun the same prompt, it comes out fine. It may sound like a conspiracy theory, but it happens too often to be a coincidence.
I've had issues occasionally, but for the most part I've usually gotten what I prompt and in those cases where I didn't, I was able to mostly fix the mistakes within Suno by extending at various points in the song, that would fix most problems aside from sound quality issues.
This is the first time where it completely ignores everything I tell it to do, from specifying the version of the model to it ignoring the instrumental setting and ignoring the genre and styles, I suppose I'll see later when I go back to my usual methods whether it's related to the audio upload feature or whether something broke while Suno was being updated for this new feature.
 
I've had issues occasionally, but for the most part I've usually gotten what I prompt and in those cases where I didn't, I was able to mostly fix the mistakes within Suno by extending at various points in the song, that would fix most problems aside from sound quality issues.
This is the first time where it completely ignores everything I tell it to do, from specifying the version of the model to it ignoring the instrumental setting and ignoring the genre and styles, I suppose I'll see later when I go back to my usual methods whether it's related to the audio upload feature or whether something broke while Suno was being updated for this new feature.
I have had issues with the extend feature where it gives a second piece as a new prompt wihen doing the extension, not just extending the current track as a whole. I may be doing something wrong. Will play with it some more.
 
Used a sound from a video I recorded of a bird tapping at my father's window today, used it in Suno's Upload Audio feature and it turned out quite well. The first 15 seconds is the actual sound from the video.

 
Last edited:
Used a sound from a video I recorded of a bird tapping at my father's window today, used it in Suno's Upload Audio feature and it turned out quite well. The first 15 seconds is the actual sound from the video.
Cool concept and cute lyrics
 
Top
Sign up to the MyBroadband newsletter
X