Audio AI GUIDE

Spleeter Stem Separation

Spleeter is an open-source tool from Deezer that splits a finished song into separate tracks (vocals, drums, bass, and more) using deep learning.

2 min readLast updated

Overview

It made high-quality stem separation fast, free, and accessible to anyone with a laptop.

Deep Dive

Spleeter, released by music-streaming company Deezer in 2019, separates a mixed recording into individual instrument stems. It ships in three pre-trained configurations: 2-stem (vocals plus accompaniment), 4-stem (vocals, drums, bass, other), and 5-stem (which adds piano). Under the hood it uses U-Net convolutional neural networks that operate on the audio's spectrogram, predicting a soft mask for each source. Multiplying the mask by the original spectrogram and inverting back to audio yields each stem. What made Spleeter famous was speed: it can separate audio roughly 100 times faster than real time on a GPU. It is widely used by DJs, remixers, transcribers, and karaoke makers, and it sparked a wave of competing separators like Demucs.

Technical Insight

Spleeter works in the time-frequency domain. Audio is converted to a magnitude spectrogram via Short-Time Fourier Transform (STFT). A U-Net (encoder-decoder with skip connections) learns, per source, a mask between 0 and 1 for every time-frequency bin. The masked spectrogram is recombined with the original mixture's phase, then an inverse STFT reconstructs the waveform. Because it estimates soft masks rather than raw audio, leakage and reused phase cause artifacts.

Strategic Impact

Access and reach

It improves accessibility through transcription, narration, and voice interfaces.

Cost and budget

Media teams can ship polished audio faster with smaller budgets.

Speed and scale

Customer-facing systems can process spoken interactions at larger scale.

The Future of Spleeter Stem Separation

Newer waveform-domain models like Demucs and hybrid transformer separators now beat Spleeter on quality, recovering crisper transients and fewer artifacts. The trend is toward higher stem counts (separating individual guitars or backing vocals), real-time on-device separation in DAWs and phones, and integration into streaming apps for instant remixing or accessibility. Spleeter itself remains a popular baseline because it is lightweight, free, and easy to run, even as research pushes phase-aware and generative approaches.

Real-World Implementation

Creating instant karaoke tracks by removing the lead vocal from a commercial song

DJs and producers isolating a drum or bass stem to build remixes and mashups

Music students extracting a single instrument line to transcribe and practice along with

Restoring or cleaning old recordings by separating and re-balancing muddy mixes

Risks & Guardrails

Voice misuse and impersonation risks increase when consent is missing.

Accuracy can drop across accents, dialects, or noisy environments.

Synthetic audio can be mistaken for authentic speech without clear labeling.

Implementation Roadmap

1

Obtain explicit consent for voice capture, cloning, and reuse.

2

Test quality across diverse speakers and background conditions.

3

Define when a human must review or approve outputs.

4

Label synthetic audio and keep provenance records for accountability.

Keep Exploring

Free newsletter

Keep up with AI in 3 minutes a day

One short email each weekday with the three AI stories that actually matter. Free forever, no ads.

One email each weekday. Unsubscribe in one click. We never sell or share your address.

Test yourself

Take the Spleeter Stem Separation quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Start quiz

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Next guide

Conv-TasNet Time-Domain Separation

Frequently asked questions

What is Spleeter Stem Separation?

Spleeter is an open-source tool from Deezer that splits a finished song into separate tracks (vocals, drums, bass, and more) using deep learning. It made high-quality stem separation fast, free, and accessible to anyone with a laptop.

Which company originally released Spleeter?

Spleeter was released as open source in 2019 by Deezer, the French music-streaming company.

What does Spleeter's 4-stem model separate audio into?

The 4-stem configuration outputs vocals, drums, bass, and an 'other' stem for everything else.

What neural network architecture does Spleeter use?

Spleeter uses U-Net convolutional networks that predict masks on the audio's spectrogram.

How does Spleeter actually extract a single source from the mix?

The network predicts a per-source mask (values 0 to 1) that is multiplied with the mixture spectrogram.

Which is a key practical advantage that made Spleeter popular?

Spleeter can separate audio around 100x faster than real time on a GPU, making it fast and accessible.