What this free karaoke maker creates
Song Separator makes an instrumental audio backing track. It does not produce a lyric video or automatically synchronized words. Standard mode splits your file into Vocals, Drums, Bass and Other. Once you mute Vocals, the mixer plays the instrumental parts together.
Choose Download Current Mix to export the exact audible mixer selection. You don't need to download and combine four separate WAV files in another program. If you'd like quieter drums or stronger bass for a practice track, change the volume sliders before exporting. Solo overrides mute while active, so make sure Vocals isn't soloed when creating an instrumental.
AI separation versus old-fashioned vocal cancellation
Traditional center-channel cancellation subtracts shared left/right audio to reduce sound mixed equally into both speakers. It may reduce a centered lead singer, but also removes other centered sounds—often bass, kick drum or snare. AI source separation instead predicts the vocal and other instrument signals individually.
The AI approach can retain more of the backing music, but it isn't perfect. Heavy reverb, layered harmonies, distortion and dense arrangements may leave traces of singing, or cause watery artifacts. Listen before saving, then decide whether the result works for your practice or performance.
What can you do with a vocal-free backing track?
- Rehearse for karaoke or an audition: sing against the instrumental audio.
- Practice an instrument: mute or solo Bass or Drums to study an arrangement or replace that part yourself.
- Adjust the balance: turn down the drums, bass or remaining instruments and export the personalized accompaniment.
- Compare to the original: use the Original reference track to hear what changed, but mute it again before exporting an instrumental.
Removing vocals doesn't grant permission to redistribute a copyrighted song. Use music you own or have rights to process and share.
Audio quality, privacy and browser requirements
Your output is a stereo 16-bit WAV at 44.1 kHz, not an MP3 or karaoke video. Your source audio is decoded, separated and mixed on your device; it is not uploaded to Song Separator for processing. A model file downloads to your browser, which must support WebGPU and have sufficient graphics memory.
Chrome or Edge on a compatible desktop computer is recommended. Safari on a Mac requires macOS Tahoe 26 or newer for WebGPU. Mobile browsers may lack sufficient GPU memory. If you get an error, see our troubleshooting guide.