New utility for separating audio in samples

[v-e-r-y]

Stumbled across this whilst reading the news this morning, from what I can gather it is a new program that separates audio in tracks into all the individual instruments including vocals.

I have not tried the software but there are possibilities here, for instance that sample you have with the drums and vocals you can't separate using existing techniques this software "seems" to do it all for you automatically.

Just thought I would share with everyone.

http://www.bbc.co.uk/news/technology-12510702

[cartesia]

Person who wrote that article obviously doesn't know what they're on about...

ce*L*emony's melodyne (yeah they got the company name wrong in the article.. great research!) DNA was there first, and does this.. and is the established industry standard...

sound forge does not do this..

hit n mix is a decidedly lame name too...

[Sinapse]

the demo was pretty superficial but all that shit works on the idea of frequency separation but all the diff sounds have frequency overlap so it never really works perfectly...may do a great job on songs with really clean separation of sounds in the frequency spectrum but as soon as you get something that is busy it's gonna be a trainwreck

[Nothingless1987]

may do a great job on songs with really clean separation of sounds in the frequency spectrum but as soon as you get something that is busy it's gonna be a trainwreck

exactly what i thought.

[cartesia]

the demo was pretty superficial but all that shit works on the idea of frequency separation but all the diff sounds have frequency overlap so it never really works perfectly...may do a great job on songs with really clean separation of sounds in the frequency spectrum but as soon as you get something that is busy it's gonna be a trainwreck

Hey man it's a little more complex than that nowadays..(at least melodyne DNA is) they still can't exactly separate the texture of an instrument but they can seem to get the attack portion pretty accurate and a bit of it the body texture a bit better than they could before...

[Sinapse]

Hey man it's a little more complex than that nowadays..(at least melodyne DNA is) they still can't exactly separate the texture of an instrument but they can seem to get the attack portion pretty accurate and a bit of it the body texture a bit better than they could before...

I'd like to read a lil more about it (prolly can't tho since it's proprietary) but just thinking bout physics of sound I wonder what you could separate it based on other than frequency...I guess you could do it based on harmonics, but diff harmonics alter one another so it's not like you could program a "trumpet harmonic" and just pull it out...idk prolly getting too technical and I really don't know the details of how it may work but I just wonder what the actual science could be besides separating by pitch/frequency...

[a.D.]

That's cheating anyway,lol...unless you can pull a Jake One & stack enough money to buy Stax's entire archive of multi-track studio sessions....

[Headphones]

But how many times have we wanted something that removed vocals from samples, or anything we didn't want from a track. Just think, if I wanted the acapella from any track, this is a way to get it! The only problem I see is it using it's own .rip format, but as long as that can be changed back to mp3 or .wav, then who cares? It's going to open pandora's box of mashups and rare remixes, that's for sure, but that's been happening lately anyway.

[tape smoke]

you can pretty much do this with adobe audition's spectral analasys, just highlight the frequency range you want, copy and paste to new file, its kinda sketchy but it works , sort of . it works better than melodyne i think

[cartesia]

That's only taking the frequency range though. not exactly the same as separating the transients/etc.. i dunno may

[cartesia]

*why is there no delete post button?*