Stable Audio represents the cutting-edge audio generation research by Stability AI’s generative audio research lab, Harmonai. We continue to improve our model architectures, datasets, and training procedures to improve output quality, controllability, inference speed, and output length.
Stability AI introduces their AI model for creating audio and music. You can try it out now: https://www.stableaudio.com/
Can this kind of stuff be used for music restoration? I have a song that sounds kind of garbled and poorly balanced (it came from RealAudio over dialup) and to my knowledge it’s never been released.
Contrary to what the other guy is talking about, you can totally use AI for restoration, or at least upgrading the audio.
However this one might not be able to do it. It depends on how this thing is programmed and if it supports something like stable diffusions control net
Totally. AI does serious improvement to visual resolution upscaling, so it should be capable of serious audio resolution upscaling. Upscaled media never is exactly like the original, but its quality is massively improved to the way humans perceive things.
you can’t really use AI for restoration at all, modern AI systems are based around generating just, whatever. and the hope is that with enough training the whatever will bias towards what you want.
this means that it can’t restore, it can’t reason about what was originally there, it can only make it’s own thing and we hope that with enough training it’ll fool humans enough not to matter.
if all you want is something to sound “better” and don’t care about the restoration aspect then sure. if you want what was originally there, no
nah, conservation would be keeping that original recording in a way as close as possible, Restoration would be restoring what was originally there.
What AI does is neither, it imagines details that never existed to make something new. It’s important that we don’t erode the meaning here because what AI does is cool, but it also does something new that does not care about the original thing.
Can this kind of stuff be used for music restoration? I have a song that sounds kind of garbled and poorly balanced (it came from RealAudio over dialup) and to my knowledge it’s never been released.
Contrary to what the other guy is talking about, you can totally use AI for restoration, or at least upgrading the audio.
However this one might not be able to do it. It depends on how this thing is programmed and if it supports something like stable diffusions control net
Totally. AI does serious improvement to visual resolution upscaling, so it should be capable of serious audio resolution upscaling. Upscaled media never is exactly like the original, but its quality is massively improved to the way humans perceive things.
you can’t really use AI for restoration at all, modern AI systems are based around generating just, whatever. and the hope is that with enough training the whatever will bias towards what you want.
this means that it can’t restore, it can’t reason about what was originally there, it can only make it’s own thing and we hope that with enough training it’ll fool humans enough not to matter.
if all you want is something to sound “better” and don’t care about the restoration aspect then sure. if you want what was originally there, no
You’re making a conservation vs restoration argument here.
Restoration is what the guy is asking about, I suspect, not conservation.
nah, conservation would be keeping that original recording in a way as close as possible, Restoration would be restoring what was originally there.
What AI does is neither, it imagines details that never existed to make something new. It’s important that we don’t erode the meaning here because what AI does is cool, but it also does something new that does not care about the original thing.
deleted by creator