Splitting songs into their components is getting progressively easier. In the past, in order to access individual vocal and accompaniment tracks of a song, you either had to purchase the original recordings from the rights holder or try to split the track yourself using sophisticated software designed for professional use.

Now the average user can efficiently extract vocal and instrumental stems in a matter of seconds and without much effort. Thanks to advances in machine learning and artificial intelligence, many solutions have emerged on the market offering high stem separation precision combined with a user-friendly interface and overall ease of use.

PhonicMind and LALAL.AI are some of the most popular AI-based services for splitting audio tracks. Both have good online reviews and at a glance, they look and function almost identically. So which service is better? How are they different? Which one should you choose? Let’s take a closer look at the splitters to help you decide.


Both LALAL.AI and PhonicMind are online services and can be used on desktop and mobile. Users of any operating systems can access the splitters without having to register, install software, etc. as long as their device has a browser and Internet connection.


Free samples or previews of separated stems are provided by both PhonicMind and LALAL.AI.

1 song - PhonicMind - $3.99 / LALAL.AI - Free (up to 3 free tracks)

10 tracks - PhonicMind - $19.99 / LALAL.AI - $10

30 tracks - PhonicMind - None (the most extreme pack peaks at 20 songs) / LALAL.AI - $30

Over 30 tracks - PhonicMind - None / LALAL.AI - Custom pricing

Upload Limits

The sites of both services are quite easy to navigate, with an upload button right on the first page. PhonicMind allows you to upload files up to 100MB in size with song length up to 9 minutes, whereas LALAL.AI’s maximum upload size is a whopping 5GB with no song duration limit. LALAL.AI’s premium packages also extend the upload size up to 150GB.

Supported Formats

You can upload songs as MP3, AAC, WMA, FLAC, WAV, and AIFF files to PhonicMind and download the isolated stems in either FLAC (if you input lossless audio) or MP3 (if you input compressed audio formats such as MP3 and WMA).

LALAL.AI supports all popular audio formats, lossy and lossless alike. As for the output formats, what you give is what you get — when you upload an audio file in FLAC, you receive separated stems in FLAC; when you upload a song in WAV, you get its stems in WAV, and so on.

Stem Separation

PhonicMind splits songs into 4 tracks — vocals, drums, bass, and other instruments/sounds. The volume of each stem can be adjusted after splitting. LALAL.AI, on the other hand, separates songs into two stems, vocal and instrumental, and has an audio processing filter for artifact elimination.

Both services use artificial intelligence for music source separation. LALAL.AI has two neural networks, Rocknet and Cassiopeia, you can select between them before uploading a file. PhonicMind has one neural network and not much is known about it.

Quality Comparison

One video demonstrating the track separation results of both services is worth a thousand words. Check out how LALAL.AI and PhonicMind extracted vocal and instrumental tracks out of the same song: