Home

Isolate Vocals Using AI with Lalal.ai

By Phil South
Happy Young Woman With Headphones

Vocal removal from already-mixed music used to be a very tight niche occupied solely by musicians and remix artists. Of course, like everything in this media-saturated world we currently reside in, now everyone can do it.

With AI music software, loops and beats creation all now being accessible to the public, this is the new mainstream, especially as the channels through which people can broadcast their work are cheap or free. Budding singers can strip the original vocalist from a popular tune and record themselves singing the classic, and musicians can have the original artist singing over their original interpretation of the music. Youtubers will fill the airwaves with cover versions.

For fun or for art, like it or not, AI software which processes the art of others and turns it into something new is big business.

This is a sponsored article and was made possible by lalal.ai. The actual contents and opinions are the sole views of the author who maintains editorial independence even when a post is sponsored.

Vocal Extraction

Lalal.ai (which I’m assuming is pronounced “lala-lie”) is an online vocal removal web app which uses AI to delicately separate the vocal track from the music in any piece of music you feed it. It then presents this to you as a pair of tracks which you can download. One file has the vocalist on their own and the other file is the music without the original singer.

Lalalai Main Interface

The online interface is very simple: upload a file and then the AI processes the music and extracts the vocal. You get the processed files available as downloads with a notation of the price for the service.

It’s a very simple tool, and although there are some small tweaks you can make to the process, there are no fine-tuning tools for the effect.

Does it work?

Software of this kind has existed for decades, in some form or another. What usually happens is that the software takes out of the middle of the stereo image, which is where the voice usually sits in the mix, and pares it off into another file. This is a fairly mechanical process but was also kind of fiddly and difficult to do.

The difference with this new tool is that it doesn’t just chop the vocal frequencies out of the middle – it listens to the voice and the surrounding music and separates it intelligently. And it does this in an easy and user-friendly way.

Lalalai Loading

Obviously, it’s not perfect, and one size doesn’t fit all. I tried it with lots of tracks, modern and classic, and even some home-brewed tunes of my own, and the results were good but variable. Obviously, some vocals are so intertwined with the music in terms of mix and frequency that it’s impossible to separate them after the fact. However, the results are impressive and at the very least interesting, and they will make for some delightful remixes and fun musical experiments.

Lalalai Processing

When you upload the file, there is a selection for the strength of the processing: either Mild, Normal or Aggressive. If you need to tweak the effect, then this is your only control. You can also turn on or off push notifications, but I’m not sure that this is of much value, as the processing really doesn’t take that long, so I mostly left it off.

Lalalai Finished 2

I imagine in the future that more controls may be added in some kind of pro version which lets you – under the hood of the AI – target the processing a little bit more directly on the voice. That’s purely speculation on my part, but as nice as this software is, it did leave me wanting a bit more scope for twiddling with the result. But for now, this simple interface makes vocal removal easy and fun for everyone.

Sounds good

Lalal.ai is available online, and the fees are based on the number of tracks and minutes you use, so 3 tracks of up to 30 mins is $5, 10 tracks of up to 90 mins is $10, and 30 tracks of up to 500 minutes is $30. It also has undisclosed rates for more massive amounts of time and tracks, which leads me to believe they have this tech as an API that you can buy for your site or software. Contact them for more details about that.

I liked this very much, and it’s a fun toy and a useful tool for low- to medium-end vocal removal. I look forward with some interest to the kinds of music this will make possible.

Subscribe to our newsletter!

Our latest tutorials delivered straight to your inbox

Phil South Avatar
Phil South
Phil South has been writing about tech subjects for over 30 years. Starting out with Your Sinclair magazine in the 80s, and then MacUser and Computer Shopper. He’s designed user interfaces for groundbreaking music software, been the technical editor on film making and visual effects books for Elsevier, and helped create the MTE YouTube Channel. He lives and works in South Wales, UK.

Read next

Roughly 99 per cent of international internet traffic runs through about 550 undersea cables — and a handful of private companies including Google, Meta and Amazon now own or lease a growing share of them
Content and cloud networks now account for three-quarters of used international bandwidth, while Google, Meta, Microsoft and Amazon own, co-own or buy capacity on a growing share of the 600-plus submarine cable systems carrying over 99 per cent of data between countries, a private map hidden below the wireless internet
Wi-Fi doesn't actually slow down when a café gets crowded — the signal is fine, but every device has to wait its turn to speak on the same frequency, and the polite queuing is what you feel as lag
Wi-Fi doesn’t actually slow down when a café gets crowded — the signal is fine, but every device has to wait its turn to speak on the same frequency, and the polite queuing is what you feel as lag
In 1988, a 23-year-old Cornell graduate student named Robert Morris released a self-replicating program to measure the size of the internet — within hours it had crashed roughly 6,000 of the 60,000 machines then online, and computer security became a field
In 1988, a 23-year-old Cornell graduate student named Robert Morris released a self-replicating program to measure the size of the internet — within hours it had crashed roughly 6,000 of the 60,000 machines then online, and computer security became a field
In July 2026, Psy’s Gangnam Style passed six billion YouTube views, fourteen years after its horse-riding dance became the platform’s first billion-view video and long after engineers had replaced a counter designed to stop at 2,147,483,647
Intricate network of tree roots and moss on a forest hillside, showcasing nature's resilience.
Suzanne Simard sealed paper birch and Douglas fir seedlings inside plastic bags, fed them carbon-14 and carbon-13 dioxide, and nine days later found carbon had crossed between species through fungal threads in the British Columbia soil beneath her boots
Close-up of glowing jellyfish swimming gracefully in deep green ocean waters.
A species of jellyfish called Turritopsis dohrnii can revert its adult cells back to a juvenile polyp stage when injured or starving, effectively restarting its life cycle, and biologists have so far failed to identify any natural limit to how many times it can do this.
Elderly man with beard and bandana, reacting to smartphone while seated indoors.
A Japanese man named Jiroemon Kimura, who lived to 116, was born in 1897 when Queen Victoria still ruled and died in 2013, meaning a single human life personally overlapped with the invention of the airplane, the atomic bomb, the internet, and Instagram
A lively view of Hollywood Boulevard with iconic landmarks and busy street life under a clear sky.
The Hollywood sign originally read HOLLYWOODLAND when it was built in 1923 as a real estate advertisement for a housing development, and it was only meant to stand for 18 months, but nobody ever got around to taking it down and the city eventually adopted it as a landmark