LALAL.ai Track Splitting Software Review (hands-on)

The LALAL.AI software is an audio extraction tool that can take music files and split them up by track, extracting vocals, and/or multiple instruments. You can then export them as an isolated MP3 file. To some degree, the quality will depend on that of your original MP3 file though you can also choose how aggressive you want the software to process your files (more on that below).

I tested the software with a full account, which basically just gives you way more upload time, allowing you to process more files.

So for this review I tested a bunch of different sounds and types of music.

I'll detail my experience in this LALAL.AI review.

Note that the company behind this software is first and foremost a software-focused group, with a specialty in signal processing.

They focus specifically on helping musicians, audio engineers, and streamers, along with a wide range of additional professions. Here's a quick excerpt from their about page.

Starting as a 2-stem splitter, LALAL.AI has grown significantly during 2021. In addition to vocal and instrumental, the service was enhanced with the capabilities to extract musical instruments – drums, bass, acoustic guitar, electric guitar, piano, and synthesizer. As a result of this upgrade, LALAL.AI became the world’s first 8-stem splitter in the same year 

IDEAL FOR: Recording, audio engineers, studio projects, post-production editing, editing existing songs, learning cover songs, and vocal or instrument students.

Note that we do not provide category ratings for audio production software.

Sound Quality Overview

The most typical split is between vocal and instrument tracks, which you can then download as separate files. This split is pretty straightforward and usually cleaner then when you go after single instruments.

Here's a full list of what you can split and export or what they call Stem Separation Files:

• Vocal and Instrument
• Electric Guitar
• Acoustic Guitar
• Bass
• Drums
• Voice and Noise
• Piano
• Synth

The quality will depend largely on the type of song you want to split up. I noticed that the heavier songs seemed far more problematic for the software, while simpler songs - particularly female vocals - did a lot better with separating vocals from instruments.

Isolated instruments behave in a similar manner. For example, acoustic guitar tracks sound better than heavy and distorted electric guitars.

I can't post exported results here for copyright reasons, but I'll list some of the vocals and bands I tested.

Female Vocals

• Taylor Swift
• Christina Aguilera
• Ella Fitzgerald
• Aretha Franklin
• Adele

I had a little bit of residual noise with the newer songs, particularly Aguilera. Oddly enough, the older tracks from Fitzgerald and Franklin were cleaner and sounded really pristine.

Male Vocals

It seems as though male vocals are more difficult for the algorithm to solve, perhaps because they're a lower frequency range that can sometimes meld with bass and/or guitar.

Here are some of the male vocalists I split up:

• Chester Bennington
• Frank Sinatra
• Maynard James Keenan
• Chris Cornell
• Kurt Cobain
• Elvis Presley
• Freddie Mercury

While testing some rock vocal exports, it's pretty clear that those are harder to dissect, especially the newer tunes. Older rock songs like "Jailhouse Rock" by Elvis actually sounded pretty decent on the vocal side. 

Frank Sinatra sounded nearly perfect.

So what I'm concluding is that heavy guitar and bass seems to confuse the algorithm a bit between what's vocal and what's instrument.

That said, most of what you get out of these tracks sound really good. 

If you can, look for songs with lots of frequency distinction between vocals and instruments. I'll give you a better result.

Splitting Up Instrument Tracks

You can take out an entire instrument track which usually comes out really clean, regardless of musical style. In some cases I caught hints of vocals in the instrument tracks, but those can be cleaned out with the LALAL.AI Voice Cleaner or additional editing with another audio app.  

When you take out single instruments, drums and piano seem to come out the cleanest.

Guitar, bass, and vocals have some crossover, but the algorithm still does a good job with these, particularly the acoustic guitar.

As I mentioned earlier, the distorted guitar will give you mixed results.

Though it really just depends on the song.

Options and Control

There are two basic elements to the LALAL.AI interface:

1. File selection
2. Processing aggression selection

All told, it's a very simple piece of software on the front end - easy to operate. 

The file selection is pretty straightforward. Use a folder/file search or just drag and drop the audio file you want to process.

The processing level gives you the following three options: 

1. Mild
2. Normal
3. Aggressive

Aggressive will work to get you a cleaner result of whatever element you isolate. Vocals - for example - will have more vocal only without as much audio leakage from other instruments.

This might make for a more pure track, but then you also risk losing more of the vocal track that you ultimately want to keep.

On the mild side, you're likely to keep all of the vocal track but also let some others slip through.

For example - when set on mild - some of my vocal splits picked up more guitar and piano as well.

Again, it just depends on the track in question.

Voice Cleaner

The voice cleaner is a second processor that helps you get excess noise out of your recordings, particularly to isolate voice or vocals.

You can checkout the voice cleaner here:

Visit the page: Voice Cleaner

Run your first split and then use the voice cleaner to further "touch up" the file you've already generated. This is particularly helpful if you have audio of someone talking in an open room with a lot of background noise.

It really helps clear out everything that's not just the spoken word from the main singer or speaker.

Pricing

Pricing for the software - which is entirely browser based - is provided in two tiers, based on volume (minutes of audio):

• Standard volume
• High volume

Here's what you have available from the standard volume tier:

• Starter at $0 - 10 minutes
• Plus at $25 - 300 or 350 minutes
• Lite at $15 - 90 minutes

All of these options are one-time fees, which basically means you'll pay as you go. This is a great value and helpful option if you don't plan to use the software forever. If you need more minutes, you can always buy another package. But, either way, you won't have a recurring membership to worry about.

Here are the higher-volume price tiers:

Value

If LALAL.AI added some additional post-processing tools, it would really their product's potential and make it worth paying more than they currently ask. But in its current state, having a one-time fee option is fantastic and a very reasonable price considering it will help people that only want to use this software for a short period of time.

Uses and Best Fit for the LALAL Software

This is great for people wanting to set up karaoke tracks, make YouTube videos, analyze tracks for cover songs, or even creating backing tracks for various projects. It has usefulness in the studio too, even if it's not to the level of isolated channels. 

Learning cover songs and helping students listen to isolated tracks might be one of the best uses of this software.

So students themselves or even regular guitar teachers should consider it.

Review Conclusion

To summarize:

LALAL.AI is a well-programmed, useful piece of software that's easy to afford. If you're in a place where you can use software like this - perhaps one of the scenarios I mentioned above - we recommend LALAL.AI confidently, especially for isolating female vocals and parsing older tracks.

Though again, it can handle most any kind of music or spoken word.

Tinker around with the free 10 minutes and run some vocal and/or instrument exports. And if you have any question about my LALAL.AI review or my review process, feel free to reach out via the comments section below.

I'll jump in and help out as much as possible.

Related Articles