is a next-generation music separation service for quick and precise stems extraction. It uses a machine-learning algorithm to precisely identify and carefully separate vocal and instrumental tracks.


Artificial Intelligence has previously been used as a solution for the complex task of vocal track isolation, but LALAL.AI is the first service to surpass even Spleeter by Deezer and PhonicMind in accessibility and quality of results.

We’ve compared all available audio splitting methods and proved that ours is the best. For the sake of objectivity, we have improved some of these methods using our music database — the studio-quality multi-track recordings (the material that sound engineers operate with).

200GB of training data — this number was artificially increased during the training up to a whopping 20TB. 45 million neural network parameters and advanced digital signal processing. All this in order to use the power of AI to improve the lives of millions of people — DJs, producers, vocalists, dancers, karaoke lovers.

The service is available starting from $5 USD for 3 input tracks or 30 minutes in total. You can also test the audio splitting functionality for free by uploading any song to the website (all audio formats supported).

More information: