Meta's SAM bot keeps 'em separated as it isolates voices and instruments from audio clips Want to hear just the guitar riff from a song? How about cutting out the train noise from a voice recording? Meta says its new SAM Audio model can separate and edit sounds using simple prompts, cutting down on the manual work typical of audio-editing tools. The release of the Segment Anything Model (SAM) Audio follows the …
I mean, having something that could create music-minus-one recordings to practice guitar to (as well as isolating the guitar part as a learning tool) would be quite helpful -- I've wanted something like this for some time now.
But it's Meta...and the Ewww factor is just too hard to ignore. I wonder if there are any (even semi-) reputable outfits that could provide this kind of tool?
Hearing aids, at least good ones, already have frequency dependent amplification, and can be tuned to your particular hearing loss frequency profile. Some claim to be able (using AI of course!) to amplify particular targets such as picking out nearby conversation within a crowded environment, or music over background noise. I got some recently, at ridiculous cost, so I have much less excuse to ignore my wife and surrounding conversations in the office are more intrusive. Oh well, such is progress.
I trialled this technology for a new audiobook and unfortunately found it really lacking, the audio kept cutting out.
I called the support team for help and would encourage anyone else who finds this same issue to do the same.
Steps to replicate:
1) Turn on
2) Tune in
Outcome: Dropouts
A similar experience was also reported when listening to Sisters of Mercy, the Uk Goth band.
1) Tune in
2) Turn off
Outcome: Burnouts
DaVinci Resolve has a useful range of tools built-in that can help clean up poor audio. I wouldn't go so far as to say it will isolate a guitar riff from a rock track but it's good at isolating a voice from noisy backgrounds. I sometimes need to edit video from events and, occasionally, the sound tech has opened the wrong mic for the current on-camera speaker. My main recourse is to switch to the audio track from the camera's own mics (as opposed to audio taken from the sound mixing desk) - run that through the voice isolation tool (with a bit of fine tuning to the strength of correction added) and the result is adequate for YouTube, etc. I've seen others use the tool to extract usable commentary from background noise that almost made the voice inaudible.
I'm not easy with the way AI is hyped in some quarters but it has its uses.
Noise removal is not the same thing at all.
I'm surprised at the uninformed comments on here.
Stem splitters have been around a couple of years and are now in a lot of products e.g. Apple's Logic (pro audio workstation software).
Far from perfect but can split a pop tune into guitar bass drums vocal with the rest as one track
