How-to guide
How to remove vocals with Spleeter vs VocalSplit
Spleeter is Deezer's open-source vocal separator, released in 2019 and still one of the most popular free options. It runs locally if you can set up Python and TensorFlow. VocalSplit runs in any browser with no setup.
Spleeter is technically free, but running it costs you the time to install Python, pip, TensorFlow, model weights and ffmpeg — plus disk space for models and a reasonably fast CPU or GPU. VocalSplit costs 99 cents and 15 seconds.
Step-by-step
- Spleeter method — Install Python and Spleeter. Install Python 3.9+, then pip install spleeter. Download the 2-stems model on first run (about 80 MB).
- Spleeter method — Run from the command line. spleeter separate -p spleeter:2stems -o output input_song.mp3. Wait for processing — much slower than cloud separation on a laptop CPU.
- VocalSplit method — Open vocalsplit.io. Any browser. Nothing to install.
- VocalSplit method — Upload, pay, download. $0.99, 15 seconds, done. Stems are 24-bit WAV instead of Spleeter's default 16-bit.
- Decide which is right for you. Spleeter: free, private (local), but high effort and older model. VocalSplit: cheap, fast, newer model, no setup.
Tips for better results
- Spleeter's older model is noticeably dated compared to production neural separators — you will hear more bleed.
- For batch processing thousands of songs, running Spleeter on your own GPU is still the cheapest option.
- Some Spleeter forks (Spleeter GUI, Audacity plugins) make the install easier but do not improve the underlying quality.
Try VocalSplit free
Upload a song and get clean vocals and instrumental stems in under 15 seconds. First split is $0.99.
Split a song