Music Demixer

Vocal remover and extractor. Music separator and stem splitter. Instrumental breakdowns. Music transcription and MIDI. Private and offline. Advanced AI technology.

Support & legal

Customer support

For account management such as refunds, cancellations, and switching between monthly and yearly billing, use this link.

For customer service, please contact us by e-mail.

Privacy policy

Our product runs privately on your computer or device. Your audio and music files never leave your computer.

When you pay for a PRO model, you get private and local access to do unlimited demixing on your device!

Some anonymous website usage data is collected to improve the service. This data is not shared with third parties.

The stem separation AI on this site uses the Demucs Hybrid Transformer model. ⚠️ You acknowledge that your usage of the stem outputs of the Music Demixer apps are governed by the Meta AI Research Academic licenses ([1], [2])!

FAQs

Q: Which browsers are supported?

Answer: For the best results, we recommend Google Chrome for the latest features and best AI processing speed. However, all modern major browsers should work, including Safari, Edge, Firefox, and Chrome.

Q: Are GPUs supported?

Answer: Yes, we support WebGPU for all modern browsers for our stem separation AI models. This will automatically use your GPU (graphics card) if available for dramatically faster demixing - measured up 10-18x faster at all quality settings. If your browser does not support WebGPU, we will use the CPU instead.

Read more about WebGPU on the Chrome developer site, and find out if your browser supports WebGPU on the Can I Use website.

Q: How much memory do I pick?

Answer:

If your browser supports WebGPU you will only have a max of 16 GB available because your GPU or video card will be doing most of the processing.

For non-WebGPU support, you can choose 4, 8, 16, or 32 GB of memory. These correspond to 1, 2, 4, and 8 threads. More threads means faster demixing. Choose the max that can fit in your device RAM/memory. Read our guide on how to pick max memory for your computer. Note that when demixing a large song, you need to give the job more memory or else risk crashing!

Q: Why are the returned audio files always WAV 44100 Hz 16-bit?

Answers:

  • .wav format: it is a simple way to directly return the output of the AI model without compression or affecting audio quality
  • 44100 Hz sample rate: it is the only one supported by the AI model, and input audio is converted to 44100 Hz before processing, which we return directly instead of resampling to the original sample rate
  • 16-bit depth: (16-bit PCM, pcm_s16le) keeps the wav files smaller and reduces memory usage of our website. However, we have added a new feature in “Advanced settings” to select 32-bit output (32-bit float PCM, pcm_f32le). Try it out and see!.

Q: How do I debug an error if the app fails or crashes?

Answer:

  1. Check the developer console: Open the developer console in your browser and look for any error messages in the console tab. Use this guide for instructions on different browsers and operating systems
  2. Check the network tab: In the developer console, go to the network tab and look for any failed requests
  3. Retry the demixing job with a different memory setting or a lower quality setting
  4. If it still doesn’t work, email us). Include the full copy-paste of the developer console output and the job settings (including audio file size and length and demixer settings)

For other questions not covered by this FAQ, contact us via email or through our Instagram account