Spleeter
Deezer's Python library and CLI with pretrained TensorFlow models that split songs into 2, 4 or 5 stems.
Pretrained 2, 4 and 5 stem models you can run from the command line or Python.
Spleeter is Deezer's source separation library that splits a track into vocals and accompaniment, four stems, or five stems. It suits developers and DJs who want a fast, scriptable separator and accept older TensorFlow-era packaging.
What it is
You install it with pip and run spleeter separate with a model such as spleeter:2stems, which writes vocals.wav and accompaniment.wav. The 4 stems model adds drums and bass, and the 5 stems model adds piano. The README says it can separate to 4 stems 100x faster than real-time on a GPU, and it can also be used to train your own model if you have isolated-source data.
Who it's for
It suits people who need quick vocal removal or want to embed separation in a Python pipeline. It is a weaker choice for the highest separation quality: the Demucs README lists Spleeter at 5.9 overall SDR against 9.0 for its own latest model.
Key features
2, 4 and 5 stem models
Separate vocals, drums, bass, piano and other with pretrained models.
Very fast on GPU
The README claims 4-stem separation 100x faster than real-time on a GPU.
CLI and Python API
Run spleeter separate or call the Separator class in your own code.
Model training support
Train a new separation model if you have a dataset of isolated sources.
Docker image
A published Docker image avoids local dependency setup.
Google Colab notebook
Try it in a browser without installing anything.
Up and running in 3 steps
Our honest verdict
What's great
- MIT-licensed code with pretrained models included
- Fast enough to batch whole libraries of songs
- Already built into commercial tools such as iZotope RX 8 and VirtualDJ, according to the README
- Simple two-line workflow: install and run one command
Where it falls short
- It is built on TensorFlow, and the README warns of known issues on Apple M1 chips and says conda is no longer recommended
- You need to install ffmpeg and libsndfile yourself, and the 2.1.0 release changed CLI options and dropped the dedicated GPU package
- Separation quality trails newer models such as Hybrid Transformer Demucs
- Deezer does not host or support GUI forks, and the README points to a separate paid product, Spleeter Pro, for professional support
Pricing & license
Free under the MIT licence; the cost is running it on your own computer. The README points to a commercial Spleeter Pro at https://www.deezer-techservices.com/solutions/spleeter/ but shows no price.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is Spleeter free for commercial use?
The README says the code is MIT-licensed. It also warns that you must get authorisation from rights owners if you process copyrighted material.
Is there a paid version?
The README links a commercial Spleeter Pro from Deezer Tech Services for faster processing and support. No price is given in the README.
Can I try it without installing?
Yes, the README links a Google Colab notebook.
