Synth Forum

Notifications
Clear all

New, Free AI Software Can Split Audio Into Separate Instruments

3 Posts
3 Users
0 Likes
1,458 Views
Michael Trigoboff
Posts: 0
Honorable Member
Topic starter
 

A few weeks ago, a company named Deezer released spleeter: free, open-source artificial intelligence-based software that can split an audio recording of a band into separate instruments, including vocals.

Here is their press release.

You need a relatively modern computer, with an Intel chip that supports some version of AVX.

If your computer doesn't support the required features, there is a free, online web-based version of spleeter you can try.

I managed to get it installed on one of my computers. It's not easily done, but I can give more details of how to do it if anyone is interested.

I've attached an example of what it can do. I had it extract the vocals of Bob Dylan singing Leopard Skin Pillbox Hat.

Attached files

leopardskin_vocals.zip (766.2 KB) 

 
Posted : 20/11/2019 2:33 am
Jason
Posts: 7897
Illustrious Member
 

"Relatively modern" means something different to different audiences. Fairly proportional to your age, generally. So to spell this out:

AVX was introduced in 2011 (Q1 for Intel, Q3 for AMD). Codename for the Intel's first CPU with AVX is "Sandy Bridge" and AMD's first is "Jaguar".

A more marketing-friendly name for the Intel "Sandy Bridge" would be Intel Core i3/i5/i7 2nd Generation. And the full marketing name for one example CPU of this generation would be "Core i7 2600".

As of today, Intel is on the 10th generation - so there are "lots" of generations of past Intel chips that support AVX. If your PC is running Win 10 - then you should be fine.

I can't speak to Mac specifics. Macs manufactured in 2011 or later would all support AVX.

 
Posted : 20/11/2019 6:54 am
Lex
 Lex
Posts: 0
Eminent Member
 

I can confirm that Spleeter works in CPU mode on my AMD Ryzen 7, so Intel is not a requirement.

It's worth noting that Spleeter can also be run on the GPU (being that it is a neural network), which should work regardless of your CPU and will be noticeably faster especially if you have an older (but still supported) CPU. This requires a CUDA-capable Nvidia GPU with a compute capability of 3.0 or higher (roughly speaking, anything from a GTX 650 onward).

 
Posted : 20/11/2019 6:26 pm
Share:

© 2024 Yamaha Corporation of America and Yamaha Corporation. All rights reserved.    Terms of Use | Privacy Policy | Contact Us