The job this page can do, and the job it cannot
This is a music separator. It is good when a voice was recorded or sung over a music bed: a lyric video, a talking head with a song underneath, a podcast sting that is actually a track. The voice goes to the vocal stem. The bed goes to drums, bass and other.
It is not a denoiser. Traffic, a television in the next room, room tone and two people talking over each other are not a lead vocal on a mix. Those will not come out clean. If what you have is a song and you want the singer gone instead, use the vocal remover.
Video in, voice out
You do not have to convert first. Drop the video and the page reads the audio track. If you already have an MP3 of the same clip, that works too, and it uses less memory than a long video.
The download is audio only. Your editor still has the picture. Mute the camera's original audio and lay this stem under it. Export WAV if you will still edit the voice, MP3 if you only need to hear it.
When the music will not fully leave
A few cases come up every time. None of them mean the file was uploaded.
- The song's lead vocal and your voice are both voice. The model keeps both.
- A sung hook that is really a synth lives in the other stem and stays out of the vocal, which is what you want.
- A vocal chopped into the beat can split. Listen to the other stem before you throw the take away.
- Long reverb on the voice stays on the voice. That is the room, not the backing track.
- Very long videos on a phone run out of memory. Trim the clip or use a computer.
Keep the tab open on a phone
Splitting a video means decoding it and then running the separator. On a phone, leave AudioDrop on screen until the download is in your files. Switching apps is the usual reason a split never finishes.
Nothing is uploaded. There is no account. If the clip is an interview or a client file, that is the point.