Make a karaoke track from any song
The music of a song without the singer, made by an AI model in this tab. For karaoke nights, singing practice and covers.
- Nothing is uploaded
- Works offline after first use
- Open source
- No watermark · limit is your device’s memory; phones hold less
Make a karaoke backing track from any MP3, M4A or WAV: an AI model takes the lead vocal out and leaves the music, on your device. It is free, needs no account and adds no watermark, and your files are not uploaded: everything runs in your browser on your device. The first run downloads the separation model (about 67 MB) and the WebAssembly engine that runs it to your browser, and after that the tool works offline.
Done
Open the network tab. Your files never appear in it.
Add files and run the tool to see every request this tab makes.
Your files are processed on this device and never sent to a server. Reload the page to clear everything from memory.
Commercial karaoke versions exist for popular hits, but not for the song your choir picked, a local band’s single or the track from a film you love. A vocal remover makes one from the original: it hears which parts of the sound are a voice and takes them out, leaving drums, bass, guitars and keys where they were.
A worked example: a 3 minute 40 second M4A of a pop song bought on iTunes. Drop it here; the music without vocals is already chosen and saved as MP3 at 320 kbps, about 8.8 MB. The song is split into pieces of about six seconds, each is separated by the model, and they are joined again, so the backing track has exactly the original timing and length. Load it into a karaoke app, a speaker or a video with the lyrics on screen. For singing practice, Pitch changer can then move it into your key.
What remains of the voice depends on the mix. A dry lead vocal comes out almost completely; long reverb tails, doubled backing vocals and ad-libs panned like an instrument can leave a faint ghost that is easy to sing over but audible in quiet passages.
How it works
- Drop the song; the music without vocals is already chosen.
- Choose MP3 or WAV.
- Press Make karaoke track. The first run downloads the model and its engine once, then the backing track downloads.
In your browser, not on a server
Your browser’s own decoders read the song, from an audio file or a video’s soundtrack, at 44.1 kHz stereo. It is cut into pieces of about six seconds, and each piece becomes a spectrogram (a short-time Fourier transform with a 6144-point window, as the model was trained with). UVR-MDX-NET Inst HQ 3, an MDX-Net model from the Ultimate Vocal Remover project (MIT), predicts the spectrogram of the music without the voice; it runs through onnxruntime-web (MIT) on the graphics chip with WebGPU where the browser has it, or on the CPU with WebAssembly. The answer is turned back into sound, and the vocals are the song minus the music.
Compared with upload-based tools: Online vocal removers upload the whole song, queue it, cap free minutes or songs per day, and ask for an account to download WAV. Here the song never leaves your device, the limit is your device’s memory (phones hold less), and the model works offline after the first run.
Verify it yourself. Give the page a short while after your first run to save itself for offline use, then turn off Wi-Fi or switch to airplane mode and run the tool again. Or open your browser's developer tools (F12, or Cmd-Option-I on a Mac), pick the Network tab, and watch it while you convert. The panel above does the same count for you.
Questions
Will it work on a YouTube video?
It works on any audio or video file you have on your device, including a video saved from your phone. The soundtrack is separated and saved as audio; the picture is not kept.
Can I change the key for my voice?
Yes, in a second step: open the backing track in Pitch changer and move it up or down in semitones without changing the tempo.
Does it keep the backing singers?
Mostly not. The model treats every voice as vocals, so harmonies and choirs are removed with the lead. Backing vocals mixed quietly under the instruments sometimes survive.
Is the song uploaded?
No. The separation runs in this tab, on your device, and nothing about the song leaves it.
More video and audio tools
Vocal remover
Split a song into the music and the voice with an AI model running in this tab. Karaoke tracks and acapellas without uploading anything.
Trim video
Cut a clip out of a longer video without uploading it. Most cuts take seconds and keep the original quality.
Mute video
Take the sound out of a video, in seconds and without re-encoding. The video never leaves this device.
Resize video
Change a video’s resolution, from 4K down to 240p or to a width and height you choose. Resized on this device, never uploaded.
Other conversions
Guides
- How to remove vocals from a songAI vocal removers pull the voice out of almost any song now, not just tracks with a lucky stereo mix. Here is how they work, and a free browser-based one.