The free-music-demixer is an AI-driven tool designed to demix music, allowing users to isolate various instruments from a music recording into separate stems. Accessible through a web application, it offers unlimited usage without any cost.
Powered by the Open-Unmix AI model with UMX-L pretrained weights, this tool executes the demixing process directly in the user’s browser, ensuring data privacy as files are never uploaded. While it operates smoothly on computers, performance may be slower on smartphones.
Users can upload songs to the demixer and break them down into bass, drums, vocals, other, and karaoke using the AI model. The demixer also supports experimental batch demixing.
Developed and maintained by Sevag H, support for the tool is available through GitHub Sponsors or PayPal. Additionally, companies in the pro music space can advertise on the platform for targeted visibility within the music and technology community.
It’s important to note limitations on the commercial use of the demixed outputs and to exercise patience during the demixing process, as it can be CPU and memory intensive. Users are encouraged to report bugs or suggest features through the project’s GitHub issues page.
While input files can be in various audio formats, the outputs are consistently stereo WAV files at a sampling rate of 44100 Hz.
