Demucs
Open-source music source separation with state-of-the-art models.
External Tool
This open-source tool is maintained externally. View the source on GitHub to learn more or run it yourself.
Browse music tools →What's next with Demucs?
Choose how you want to get started.
Use it free
Open the official tool or demo — no account needed.
Self-host it
Run the open-source version on your own infrastructure.
What is Demucs?
Demucs is an open-source audio source separation tool developed by Facebook Research, designed to isolate individual audio tracks from mixed recordings. Its primary purpose is to decompose complex audio signals into distinct components, such as vocals, drums, bass, and other instruments. Audio engineers, music producers, and researchers use Demucs to extract clean audio elements from recordings where multiple sound sources are layered. The tool addresses the challenge of separating overlapping audio tracks, which is critical in tasks like remixing, remixing, and academic analysis of audio content. By leveraging hybrid spectrogram and waveform processing techniques, Demucs provides a balance between computational efficiency and separation quality, making it a popular choice for both professional and experimental applications.
How it works
Demucs is a machine learning-based audio source separation tool that uses hybrid spectrogram and waveform processing to isolate individual tracks from mixed audio. It was developed by Facebook Research and released under the MIT license, with a large community of contributors and users. The tool is particularly useful for separating vocals from instrumental tracks, isolating specific instruments, or extracting clean audio from recordings with multiple overlapping sound sources. Its primary use case is in music production, where clean separation of audio elements is essential for remixing or further processing. Demucs supports separation of multiple audio tracks, including vocals, drums, bass, and other instruments. It can handle both stereo and mono inputs, and its hybrid approach combines spectrogram-based models with waveform-based processing to improve accuracy. The tool also includes pre-trained models for common tasks, such as vocal separation, and allows customization through configuration files.
How to use it
- 1Clone the forked repository from github.com/adefossez/demucs. 2. Install dependencies using the provided environment files (environment-cpu.yml or environment-cuda.yml). 3. Run the separation script with a pre-trained model, specifying the input audio file and desired output tracks. 4. Use the command-line interface to process audio files, with options to adjust parameters like model version or audio normalization. Practical tips include using the GPU for faster processing, verifying model compatibility with the input audio format, and testing different pre-trained models for optimal results. The documentation and example files in the repository provide guidance for specific use cases.
What it can do
- music separation
Use cases
Assumptions and limitations
Assumptions
- source: https://github.com/facebookresearch/demucs
- license: MIT — free to use
- privacy: Self-hosted — you control your data
Limitations
- The original repository is archived and no longer actively maintained
- Limited support for real-time audio processing
- Requires technical expertise to set up and customize
- Performance may degrade with highly complex or non-standard audio inputs
- Dependence on specific hardware configurations (e.g., GPU for optimal speed)
Understanding the result
Open-source music source separation with models.
Tool details
- Clearly flagged when a network request is needed.
- No account, no sign-up, and no tracking of your content.
- Powered by (MIT).
- Built with
- (facebookresearch/demucs)
- License
- MIT
- Runs locally
- No — requires a network request
- Verification
- Not yet verified
- Input
- Query
- Output
- Text
Built with facebookresearch/demucs. OpenToolVault provides the discovery and browser interface while crediting the original project maintainers.
- Built with
- License
- MIT
Open-source project
OpenToolVault is an independent directory. We are not affiliated with or endorsed by this project.
References
- / — GitHub Repository
Upstream project · GitHub
- MIT License
Upstream project
Frequently asked
Is Demucs still actively maintained?
The original Facebook Research repository for Demucs was archived on January 1, 2025, and is now read-only. Active development has moved to a forked version hosted at github.com/adefossez/demucs. While critical bug fixes may be addressed in the fork, the project is no longer officially maintained by the original team.
How does Demucs perform audio source separation?
Demucs uses a hybrid approach combining spectrogram-based models and waveform processing. It first converts the audio into a spectrogram, which represents frequency over time, and applies a neural network to separate the spectrogram into individual tracks. The results are then converted back to waveform format. This method balances computational efficiency with separation accuracy, allowing for high-quality output while maintaining reasonable processing times.
How do I separate vocals from a music track using Demucs?
First, clone the forked repository and install dependencies. Then, use the command-line interface to run the separation script, specifying the input audio file and the desired output tracks (e.g., 'vocals' and 'instrumental'). Pre-trained models for vocal separation are available in the models directory. Adjust parameters like audio normalization and model version as needed, then process the file to generate isolated tracks.
How does Demucs compare to alternatives like Spleeter?
Demucs and Spleeter are both audio source separation tools, but they differ in architecture and performance. Demucs uses a hybrid spectrogram-waveform approach, while Spleeter relies on a deep learning model trained on a large dataset of mixed audio tracks. Demucs is often praised for its high-quality separation of vocals and instruments, but it requires more computational resources. Spleeter, on the other hand, is easier to use for basic tasks but may not match Demucs' precision in complex scenarios.
What should I do if Demucs fails to process my audio file?
Common issues include incompatible audio formats, insufficient GPU memory, or outdated dependencies. Verify that the input file is in a supported format (e.g., WAV or MP3). If using a GPU, ensure it meets the minimum requirements specified in the environment files. Check for updates to the forked repository and reinstall dependencies if necessary. For specific error messages, consult the project's documentation or community forums for troubleshooting guidance.