About Extractune
Extractune is built for musicians, singers and video creators who need the sound from a video, or one part of a song, without handing their files to someone else's server.
Updated
What it does
Pick a video or a song. Extractune pulls out the audio as MP3, WAV or WebM, splits the voice from the drums, bass and everything else, and gives you one-tap acapellas, instrumentals and practice tracks. The studio lets you mix the parts, slow a song down without changing its pitch, change its key, and loop the hard bit. You can save the video back with just the voice, and get the words, and the song's name, from what's said or sung.
Who it's for
- Musicians learning a part by ear, or making a backing track without it.
- Singers making karaoke tracks and moving songs into their key.
- Video creators keeping the talking and losing the music, or turning a clip into audio.
- Anyone who wants the audio of a video they have, fast and for free.
How it works, and why it's private
Everything runs in your browser. Extraction uses the WebCodecs API. Splitting uses HTDemucs, the open-source separation model from Meta's research team (the paper, the code), running on your device through ONNX Runtime Web. So your video is never uploaded. The one exception is Get the words: when you tap it, a compressed copy of the audio is sent to be transcribed and a short clip to identify the song. The privacy page has the details.
Free, with no account
Because the work happens on your device, Extractune can be free without ads or sign-ups. See what's included and the limits.
Contact
Questions, bugs or ideas: support@extractune.com.