Demucs
Open-source model that splits songs into vocal, drum, bass and other stems.
About Demucs
Demucs is an open-source music source separation model, originally from Meta AI Research, that splits a song into vocals, drums, bass and other. The htdemucs_6s model adds guitar and piano stems (piano is still weak). Version 4 uses a hybrid transformer that works on both the spectrogram and the waveform, and a fine-tuned htdemucs_ft trades four times the runtime for slightly cleaner stems. Install with pip or run via uvx; it needs about 3 to 7 GB of GPU memory, or runs on CPU at roughly 1.5x track length. Many online stem splitters run it under the hood. The author is no longer adding features.
Key Features
💰 Pricing
Free and open source (MIT license).
Frequently Asked Questions
Is Demucs free?
What is Demucs used for?
What platforms does Demucs work on?
What are the best alternatives to Demucs?
Reviews
Loading reviews...
Questions & Answers
Questions submitted will appear here soon
Join the community to share knowledge about Demucs
📊 Comparison vs Alternatives
| Feature | ||||
|---|---|---|---|---|
| Pricing | Free | Free | Freemium | Paid |
| Rating | — | ⭐ 4.4 | ⭐ 4.5 | ⭐ 5.0 |
| Free Tier | ✅ | ✅ | ✅ | ❌ |
| Popularity | 50% | 82% | 80% | 80% |
Are you the creator of Demucs?
Claim this listing to get a verified badge, edit tool details, and respond to reviews. You’ll sign in and confirm a work email so we can verify ownership.
Claim this listingQuick Stats
Categories
Demucs Alternatives
The Weekly AI Tools & Skills Digest
Launching soon: trending agent skills, new tools, and what's actually worth trying. Sign up to get the first issue.
Free forever. Unsubscribe anytime.