If you’ve been messing around with music production or anything audio-related, you’ve probably heard of stem splitting. But if not, that’s okay — this guide’s for everyone. No matter if you just want to refresh your memory or learn it all from scratch, we’ve got you covered. Plus, we’ll share some of the best stem separation software out there to help you get started.
What Is Stem Splitting?
Stem splitting is the process of separating individual parts of a multi-track audio file or a piece of music into separate elements or stems. The most typical stems are separate tracks for vocals, drums, bass, and guitar, but other instruments can be included as well. The goal is to take a fully mixed recording and separate it into individual parts, so you can use them for all sorts of music-related tasks.
Stem splitting is commonly used in the context of remixing, music production, and DJing, where the creator works with specific elements of a song while leaving others out. This is how they get to manipulate individual parts of a track without affecting the entire piece.
How Audio Stem Separation Works
Audio stem separation uses a variety of techniques, but the most popular one is source separation. This process uses algorithms to analyze an audio mix and determine where each sound or stem fits. Once they’re separated, you can manipulate each element on its own.
Source separation is mostly based on blind source separation. It uses a few different methods, but for now, the key ones to know are independent component analysis and non-negative matrix factorization, which are powered by machine learning. These techniques look at the signals, identify patterns, and figure out how the different sounds are connected.
Recently, deep learning has made a big impact on improving stem separation, so you can already take advantage of that. Algorithms like convolutional neural networks dig through an audio mix, identify unique patterns, and separate the sounds more accurately than ever. Plus, they are trained on different types of audio, so they’re flexible and are able to deliver strong results for various separation tasks.
Best Stem Separation Software
LALAL.AI
LALAL.AI is a popular stem splitter that you can use directly in your browser. It lets you isolate vocals and a variety of instruments like acoustic guitar, electric guitar, synthesizers, bass, drums, piano, strings, and wind. It processes audio quickly, depending on how many parts you're splitting, and offers an easy way to work with your stems.
LANDR Stems
Using LANDR Stems is simple, too — drag your song or mix into the plugin, and it takes care of the rest. You’ll have your stems separated and ready to work with. You can solo or mute them, and if you want to use them elsewhere, you can download the stems or pull them directly into your DAW.
Gaudio Studio
Gaudio Studio is another great example. It was once completely free (which is what initially caught people’s attention), but now it operates with a tiered pricing model ($0.04-$0.14 per minute) and offers a free trial. The trial allows up to 20 minutes of audio processing, which is still worth checking out.
You’ll need to sign up with a Google account. After that, upload the track, choose the instrument layers to isolate, and submit it for processing.
And this list is non-exhaustive. There are now plenty of other stem separation tools, both desktop and online, such as Moises, PhonicMind, Hit'n'Mix RipX, and many more.
Stem Separation in Popular Music and DJ Software
AI-powered stem splitters and vocal removers have actually been around longer than most people realize. You might even have access to one already through the music software you use.
For example, iZotope RX comes with its Music Rebalance feature, and FL Studio’s Producer Edition has a built-in stem separation tool. In 2024, Apple also brought some AI-driven features to Logic Pro 11, including their own stem splitter.
DJ software like Serato DJ, Rekordbox, and VirtualDJ also offer stem separation. But keep in mind that it’s only available in real time, which means you can’t download the separated stems for later use.
Difference Between AI-Based Stem Separation and Generative AI
AI-based stem separation and generative AI are two very different approaches, and for musicians, one is far more useful than the other.
An AI-based stem splitter is a practical tool that saves time compared to manually separating tracks. It’s not always perfect, especially when sounds overlap, but the technology is steadily improving and offers a lot of potential for musicians. As AudioShake's CEO told us in our interview, "If you want to clone Taylor Swift's voice or you want to train a music model that can process something realistic when someone types in a text prompt or a voice prompt, you need to have a concept of Taylor Swift. Stem separation doesn't work that way. We don't need to train on Elvis's vocals to be able to separate Elvis. It's uncontroversially beneficial to the artists and the rights owners because you're enabling them to take this track, split it apart and then do things with it that either are new creative opportunities or it's actual monetisation."
Generative AI, on the other hand, creates music from scratch rather than separating elements from a pre-existing track. On top of it, this tech tends to replicate elements of existing music, which might lead to copyright issues.
So, don’t get the two confused; not all AIs are built the same.
What Can You Do With Stem Separation?
Stem separation offers far more than remixing tracks or creating mashups. It works equally well in music production to film editing and sound design, and we’ll give you a few examples.
For music producers, artists, and composers
- Isolate the perfect sample — Pull out vocals, beats, or that one amazing riff to use in your next track.
- Rework songs — Break songs down into parts and put your own spin on them with remixes and new arrangements.
- Fix noisy tracks — Clean up recordings by removing unwanted sounds or fixing rough spots.
- Make karaoke tracks — Strip out vocals to turn any song into an instrumental version for karaoke.
- Prepare your music for sync licensing — composers can trailerize their music or sync agencies can use this technology to change the vocals to a track, to make them better align with the visual message.
For filmmakers and video creators
Stem separation gives you the ability to fine-tune every sound element and, for instance, tidy up dialogue:
- Separate voices — Pull dialogue away from background noise for smoother editing and dubbing.
- Clean background noise — Get rid of distracting sounds to make conversations clearer.
- Restore old audio — Improve the sound quality of older recordings by isolating and enhancing dialogue.
You can also use it for flexible post-production:
- Adjust individual tracks — Change the levels of dialogue, music, or effects separately to perfect the mix.
- Create custom sounds — Use isolated audio elements to make new effects or build up your sound library.
- Adapt music for videos — Remove vocals or instruments to create fitting background music for your project.
- Build unique scores — Use split instrumental tracks to design a custom score that matches your vision.
Why Is Stem Separation So Complicated?
At first, stem separation seems simple. You might think it’s just a matter of untangling the audio, but it’s actually much more difficult, which is why AI technology is often the best solution. Here’s why:
Mixing means losing data
When a song is mixed, all the individual stems — vocals, instruments, effects — are combined into a two-channel audio file (left and right). This process loses some details, and trying to reverse that is like putting together a torn-up letter.
Instruments compete for the same space
In a mix, sounds tend to overlap in both frequency and stereo position. For example, vocals and guitars may share similar frequencies or both sit in the center of the stereo field. Separating them without advanced tools is very difficult, even with the best technology.
Audio is full of data
Music carries a lot of data. With a standard sampling rate of 44.1 kHz, every second of audio contains 44,100 measurements. This is a lot compared to text or images, and it’s why most separation tools work with short segments.
Imperfections are easy to notice
Small flaws in audio are easy to notice. Unlike images, where one pixel might go unnoticed, even tiny mistakes in audio stand out.
The way we hear isn’t the same as audio files
Audio files store sound as waveforms, which track changes in volume over time. But our brains process music through qualities like pitch and rhythm, which are better represented visually as a spectrogram. This difference makes it harder to separate sounds in a way that matches how we hear them.
Figuring out how to label stems is tricky
For example, should a distorted guitar and a clean guitar be considered the same stem? These decisions matter, and some tools still struggle with separating more than three basic stems (vocals, bass, drums).
Every genre is different
Every genre uses different mixing techniques, instruments, and stereo placements. What works for one genre might not work for another.
If you were to try separating the stems manually, you’d run into all these challenges along the way.
Audio stem separation is an exciting area that’s growing fast, and new methods and algorithms are popping up all the time. In the future, there will surely be better deep learning techniques, and new approaches will make separation even more accurate and efficient. But for now, the solutions you have are still pretty good.
