🎚 Stems

← Studio🎤 Transcribe🎛 Multi-Track
📊 Honest about quality:
• This tool's free method: Uses center-channel cancellation. Works well when vocals are panned center and instruments are panned wider (most modern recordings). Excellent for karaoke-style instrumental tracks.
• Full ML-based stem separation (drums + bass + vocals + other isolated cleanly) requires heavy models like Demucs or Spleeter — too big to run smoothly in browser, and using them on a server requires GPU compute.
• For pro-quality separation: use one of the services below — they have free tiers for occasional use.

1. Free Browser-Based Separation (no upload to server)

🎤 Vocal Isolator / Karaoke Maker

Drop an MP3 or WAV → get an instrumental (vocals removed) AND a vocals-mostly track. Works entirely in your browser — your audio never leaves your device.

2. Pro Stem Separation (free tiers + paid services)

Moises free tier

5 separations/month free. Drums + bass + vocals + other isolated by AI. The original.

Open Moises ↗

LALAL.AI free tier

10 minutes free. High-quality vocal/instrumental, drums, bass, piano, electric guitar.

Open LALAL.AI ↗

AudioShake free tier

Free signup. Good for stems including dialogue + music separation. Stripe checkout for more.

Open AudioShake ↗

Vocal Remover (free) free

Free browser-based separation tool. No signup, no limits — but quality varies.

Open Vocal Remover ↗

3. Self-Hosted Pro Models (advanced)

Run Demucs locally (best quality, free)

Demucs from Facebook AI is the state-of-the-art free stem separator. Install Python + run from terminal:

pip install demucs
python -m demucs --two-stems=vocals "your_song.mp3"
# Produces /separated/htdemucs/your_song/vocals.wav + no_vocals.wav

4-stem (drums + bass + vocals + other) version:

python -m demucs "your_song.mp3"
# Produces 4 separate WAVs — bass, drums, vocals, other
GitHub: Demucs ↗

Spleeter (Deezer, also free + self-hosted)

Older but still useful. Faster than Demucs, smaller models.

pip install spleeter
spleeter separate -p spleeter:2stems -o output/ "your_song.mp3"
GitHub: Spleeter ↗

What to do with the stems

Once you have isolated tracks, drop them back into Multi-Track as separate tracks. Mix, mute, solo, balance — then bounce a new master to WAV. Or load just the vocals into Transcribe to convert the melody to sheet music.