F.A.Q

How does this work ?

This service relies on machine learning to separate the vocals from the instrumental. We use the open source library spleeter developped by deezer's research team.

My instrumental track sounds a bit glitchy ?

The task of separating audio from a mixed sources is a very hard one. Even using state of the art software for this task; the result often contains imperfections audible by a human ear. That's just how it is...

I use Android and my file is not accepted !

Some Android phones display files as mp3 but do not actually name the file with the proper extension. If this is the case, you may try adding ".mp3" to the file name and uploading again. If it still doesn't work, you can ask for help through the contact section (send your file).

What music does it work best on ?

Results are better with acoustic music (jazz, classical, etc...). Voices with lots of processing (autotune, distortion) deteriorate the quality of the results.

Is it free ?

Yes, up to 7 files per week (i.e. 364 per year) ! If you enjoyed this service and wish to support production and development costs and/or need to process more files this week, feel free to donate/buy credits through PayPal or credit card. Thank you !

What do you do with the music uploaded by users ?

Your music is only stored for the duration of processing. Your file is immediately deleted once the instrumental is generated.

How long does my karaoke track stay available ?

The karaoke extraction stays available for 3 hours at the address of the download page. We reserve the right to delete the file before if necessary.

Can I use the instrumental track ?

We do not hold any claim on the files we process. Make sure you have the appropriate licenses with the license holders of the original audio file if want to use an extracted karaoke.

This "spleeter" you use, is it the same as in melody.ml and moises.ai ?

Yes. As of the writing of this FAQ and to the best of our knowledge, melody.ml, acapella-extractor.com and moises.ai all use the exact same AI library to perform extraction. Results for vocal and instrumental separation should be identical.