Stem separation technology has advanced rapidly since it first appeared, with systems recognising an ever-expanding list of instruments and improving exponentially in both quality and speed. Stem separation is becoming more widely available too, with online services, plugins and standalone apps dedicated to the task, and a growing number of DAWs boasting built-in stem separation tools, with Cubase Pro 15 being just the latest to gain the ability. All of these systems work slightly differently and deliver different results, so with the second half of the decade upon us, we thought it might be an opportune moment to take stock of what’s hot - and what’s not - in the world of stem separation.
How we tested
In order to get as accurate a picture as possible, we started by testing 11 of the most popular stem separation tools, feeding the same set of three mixed songs through each and comparing the results before whittling things down to a top five. Those tools, in no particular order, are: Ableton Live 12 Suite, Steinberg Cubase Pro 15, FL Studio 2025 Producer Edition, iZotope RX 11 Standard, LALAL.AI, LANDR Stems, Apple Logic Pro 11.2, Moises.ai, Steinberg SpectraLayers Pro 12, PreSonus Studio One Pro 7, Ultimate Vocal Remover (Free and open source). The details of how each of these goes about its business are different, but all can extract vocal, drum and bass stems. Some, such as Izotope RX, then treat the remaining audio as a single “other” stem, whilst tools such as LALAL.AI and Logic Pro can go deeper and extract guitars, pianos and more.
The depth of processing - i.e. how much time the system spends analysing and processing the audio - has a major bearing on the quality of separation. Fast settings sacrifice accuracy for speed, whilst slower processing will yield better results albeit whilst leaving you twiddling your thumbs for a while. Not all of the systems tested allow the processing depth to be changed, but with those that do we always tested using the highest setting. The only exception to this was UVR’s Demucs mode, whose maximum settings were too much for our test machine to handle.
Want all the hottest music and gear news, reviews, deals, features and more, direct to your inbox? Sign up here. We relied mainly on listening tests when evaluating the results, but also took into account how versatile a system is. This is a subjective measure taking in things like the number of instruments a system recognises, whether it offers additional processing tools, whether it’s easy or awkward to work with, and so on. We also performed difference testing, playing a mix of the separated stems against a phase-reversed version of the source mix, thereby revealing the accuracy of the processing (i.e. if the difference test results in silence then the stem separation is completely lossless).
The source mixes
The nature of a mix has a considerable bearing on how successfully a separator can recognise the instruments and extract the stems. Conventional acoustic instruments are more readily recognised than the hard-to-define sounds of synthesizers. And the more instruments/parts/voices there are in a mix, the harder it becomes to separate those elements cleanly. In order to create a fair basis for comparison, we chose three songs with very different instrumentation, complexity and character. Each was chosen specifically to test and reveal different aspects of the stem separation process.
- Remember Jones - Don't Put Me On Hold: fullrange instrumentation including backing vocals, guitars, brass, organ and electric piano.
- Hayley Reardon & Pau Figueres - In The Good Light: simple vocal and acoustic guitar, testing vocal extraction and the impact on the remaining audio.
- Albert Kader - Whiptails: electronic dance with synthesized parts and non-traditional timbres, challenging for AI recognition.
Not making the cut: several tools didn’t perform as well as the top five. The big-name DAWs Cubase Pro and Studio One, for instance, showed limitations in extraction quality or consistency. UVR’s Demucs mode is strong for a free tool, but overall its best results come from dedicated modes like MDX-Net for vocal extraction, which can be exceptional but hardware-dependent. The aim here is to highlight the tools delivering the cleanest, most usable stems across representative styles.
The results
- Ultimate Vocal Remover (UVR) is a free and open-source app available for Windows, macOS and Linux. It includes various processing models and exceptional control over extraction settings. If you have an Nvidia GPU, UVR can offload processing to it, speeding things up. The MDX-Net mode yields lossless vocal extraction with minimal impact on other stems, notably in In The Good Light. SCORE: 13/20.
- LANDR Stems runs as an instrument plugin inside a DAW and pulls stems back into the project. It recognises vocals, drums, bass and other stems with consistently good results, though the guitar part can lose some clarity when the vocal is separated. SCORE: 13/20.
- LALAL.AI online service offers ten separation types, with excellent instrument recognition for extended instruments, though its interface is clunky and processing is not fully lossless due to discrete stem extraction processes. SCORE: 15/20.
- Steinberg SpectraLayers Pro is a professional spectral editor with Unmix Song and other processors; it provides high customization, enabling extraction of guitar, piano, sax and brass, and even learning a specific instrument’s sound. Processing depth can be Fast, Balanced or High. SCORE: 15/20.
- Apple Logic Pro Stem Splitter (11.2) refined the feature with up to six stems: Vocals, Drums, Bass, Guitar, Piano and Other. It offers fast, accurate results, with the vocal stem extraction showing minimal impact on adjacent guitar parts. SCORE: 16/20.
Verdict: None of the tools produce perfect results; artefacts and cross-contamination between stems occur across the board. However, the best performers deliver highly usable vocal, drum and bass stems with convincing tone and minimal artefacts. UVR and Moises stand out for vocal and background vocal separation, while Logic Pro’s Stem Splitter is a strong all-rounder with excellent accuracy and clean extractions.
In-depth look at top performers
Ultimate Vocal Remover (UVR)
UVR is free, open-source, and offers many models, including Demucs and MDX-Net for vocal isolation. The Kim Vocals 2 model yields exceptionally clean vocal extractions, with smoother reverb tails preserved. The MDX-Net mode extracts vocals with minimal impact on other instruments, delivering lossless outcomes notably in In The Good Light. Running UVR on Windows is straightforward with a GUI version; macOS can be more challenging due to security constraints. If you’re GPU-equipped and comfortable with setup, UVR is the strongest free option for vocal work. SCORE: 13/20.
LANDR Stems
LANDR Stems operates inside a DAW as a plugin and processes vocals, drums, bass and other stems via LANDR’s servers. It typically produces crisp drum and vocal stems with preserved bass high-end, though guitar clarity may suffer when vocal stems are extracted. Its integration with LANDR’s library adds value for a cohesive production workflow. SCORE: 13/20.
LALAL.AI
Online service offering ten separate types of separation; excels at extended instruments like guitars and pianos, with strong overall vocal and instrumental separation. The interface and pricing model can be a hindrance for some users, and advanced features require paid tiers. SCORE: 15/20.
Steinberg SpectraLayers Pro
Powerful spectral editing with Unmix Song, Unmix Drums, Unmix Chorus, and instrument learning. It’s capable of extracting guitar, piano, sax and brass with a choice of processing depths. It can place a part in the “Other” stem if the algorithm recognizes it better there. It’s highly capable but not always perfectly accurate in instrument recognition. SCORE: 15/20.
Apple Logic Pro
Stem Splitter in Logic Pro 11.2+ offers six stems with superb recognition accuracy and clean extractions, including successful piano and guitar stems and minimal impact on adjacent parts. The fixed processing depth yields reliable results, and the integration with Logic’s workflow is a major plus for fans of the DAW. SCORE: 16/20.
A closer look at notable mentions
- Moises.ai provides four-stem separation with mobile accessibility and extra features like chord detection and practice-focused tools. It’s a strong performer for vocal extraction and background vocals, albeit with some artefacts in vocal isolation. SCORE varies by test but is highly usable for on-the-go work.
- Steinberg Cubase Pro 15 and Studio One Pro offer built-in or integrated tools but were not top performers in our tests, especially for pure stem quality against rivals.
- MXP/Other Tools include UVR’s MDX-Net modes and various free/offline tools like StemRoller and OpenVINO plugins, which provide good baseline results but may require more setup or compromise on artefacts.
Practical guidance for choosing a stem splitter
- For vocal extraction on sparse mixes, UVR with Kim Vocals 2 is hard to beat, delivering clean vocal stems with preserved reverb tails.
- For background vocals and layered harmonies, Moises often provides robust results, especially when you need ready-made practice tools and mobile access.
- For instrumentals and remixes, MDX-23xc InstVoc or GAUDIO Studio-style processing can yield clean instrumentals with minimal bleed.
- For fast, integrated DAW workflows, Logic Pro’s Stem Splitter is among the best built into a DAW, with excellent recognition in a tightly coupled environment.
- Consider a hybrid approach: use your DAW’s built-in tool for quick edits and switch to UVR or LANDR for more critical stems or when you need deeper control over processing depth and model selection.
Tables and quick-reference
| Tool | Best use | Strengths | Limitations | SCORE |
|---|---|---|---|---|
| UVR | Vocal extraction | Lossless MDX-Net mode; GPU offload | Setup complexity; macOS security issues | 13/20 |
| LANDR Stems | DAW integration | Consistent recognition; easy workflow | Guitar clarity may drop with vocal separation | 13/20 |
| LALAL.AI | Extended instruments | Strong instrument recognition; versatile | Clunky UI; paid tiers required for advanced features | 15/20 |
| SpectraLayers Pro | Professional editing | Unmixing with learning and multiple processors | Not always perfect at instrument recognition | 15/20 |
| Logic Pro Stem Splitter | Integrated DAW workflow | Fast, accurate, secure (Apple Silicon) | Fixed depth; Mac-only | 16/20 |
In practice, the best choice depends on your use case: hobbyists can rely on built-in DAW tools; professionals may opt for UVR or Moises for high-quality vocals and deeper control; heavy users may benefit from LALAL.AI for batch processing and API access. Regardless of choice, no stem splitter perfectly recreates the original stems-if you can obtain the real stems from the artist or producer, that remains the gold standard. The evolving landscape means ongoing testing and adaptation are wise if you rely on stem separation for critical projects.
I Tested Every AI Stem Splitter So You Don't Have To
Whether you’re pulling vocals for a remix, creating a karaoke track, or isolating background vocals for a mix, these tools offer powerful options. For many, a practical workflow combines DAW-native capabilities with a selective third-party tool to maximize quality while preserving speed and convenience.