MotionShine01 / Song
Drop the song
Take the singer out with AI, or keep a quiet guide voice to sing along with.
Free tool
Karaoke maker, any song, your lyrics.
Drop in a song and take the singer's voice out with AI, paste the lyrics, and tap along once to time every line. You get a karaoke video where each line fills with colour as it is sung and the next one waits below. Download an MP4 for a party, a TV or YouTube. Nothing is uploaded.
Start timing plays the song from the beginning: press Space (or the red button) the moment each line starts, and once more when the last line ends. Then use the minus and plus to nudge a line by a fifth of a second.
Song
Skip this if your file is already an instrumental. The first time downloads a 181 MB AI model, kept for next time.
Look
Everything runs in your browser. Your song and lyrics are never uploaded.
How it works
Song, words, sing.
02 / Words
Time the lyrics
Paste the lyrics and tap Space as each line starts. Nudge any line after.
03 / Party
Download the video
Lines fill with colour as they are sung, the next line shows below.
Questions
Good to know.
How good is the voice removal?
It uses HTDemucs, an AI model from Meta's research team that splits a song into voice, drums, bass and the rest. On most pop and rock songs the voice almost disappears; reverb or backing singers can leave a faint trace, which the music usually covers.
Where do I get the lyrics?
Type them, or copy them from your own sources. Lyrics are protected by copyright, so a karaoke video for a party at home is one thing; posting one publicly may need permission from the songwriters.
Why tap instead of automatic timing?
Tapping along once takes as long as the song and puts every line exactly where you feel it starts, even in songs where words are hard to hear. "Spread evenly" gives a rough start if you prefer to adjust lines by hand.
Is my song uploaded?
No. The AI model runs in this browser tab, and the video is made here too.
Which browsers work?
Current Chrome, Edge and Safari on a computer. Taking the voice out is much faster on a computer with a graphics chip that the browser can use (WebGPU).