The short answer
Open the Instrumental Maker, choose the song and let the AI remove the vocals. For a pure karaoke track press “Download instrumental”; for a track with a soft guide voice, switch “Vocals” back on at about 10 to 20% and press “Download this mix”. Both are WAV files you can play in any player or karaoke app.
Open the free toolMake a karaoke version of a song, step by step
Pick the original recording
Use the studio version in the best quality you have. A WAV, FLAC or high-bitrate MP3 gives a cleaner backing track than a live recording or a low-quality copy.
Remove the vocals
Choose the file and start the separation. The first run downloads the AI model once (171 MB); after that, a three-minute song takes well under a minute on a computer with WebGPU and several minutes without it.
Decide on a guide voice
The Instrumental Maker opens on the “Instrumental” preset, so you hear the song without vocals. If you want help with the melody, switch “Vocals” on and lower its volume.
Save the track
“Download instrumental” saves the song without any voice. “Download this mix” saves exactly what you hear, including a quiet guide voice. Both come as stereo WAV files at 44.1 kHz and 16-bit.
Which karaoke version fits your evening?
One separation gives you four stems, so you can save several versions of the same song without running the AI again. Set the switches, download, change them, download again:
| Version | Settings | Download |
|---|---|---|
| Classic karaoke track | “Instrumental” preset (vocals off) | “Download instrumental” |
| Karaoke with guide voice | Vocals on at 10–20%, all other stems at 100% | “Download this mix” |
| Practice track for a singer learning the song | Vocals at 40–60%, so you can sing along with the original | “Download this mix” |
| Drumless track for drummers | Drums off, vocals on or off as you like | “Download this mix” |
| Bassless track for bass players | Bass off, everything else on | “Download this mix” |
Why a quiet guide voice helps
A completely clean instrumental is what you want on stage. At a party, many people sing better with a hint of the original voice underneath. It tells you when to come in after a long intro, keeps you on the melody in the bridge and covers the moment when you forget a line.
At 10 to 20% the original singer stays in the background, and your voice through the microphone or the room stays in front. If you notice that you are following the recording instead of singing, lower it further. If guests are unsure of the song, raise it a little.
The guide voice also hides small leftovers of the separation. Reverb tails or a faint backing choir that stay in an instrumental are hard to hear once a soft voice is there anyway.
Changing the key and showing the lyrics
The Vocal Remover separates and mixes the song, but it does not change its key or tempo, and it does not show lyrics. If the song is too high or too low for you, load the WAV into a music app, karaoke app or audio editor with a pitch or key control and move it a semitone or two. A WAV file is a good starting point for that, because it has not been compressed again.
For lyrics, the simplest way is a printed sheet or a lyrics page on a second screen. Many karaoke apps can also display lyrics next to your own backing track. If you make a lyric video, put the WAV file at the very start of the timeline; since it is as long as the original song, the timing of any lyrics synced to the original stays correct.
Before the party: a short checklist
- Prepare the tracks at home. The model download happens once, and after that the tool works offline, but separating ten songs on a laptop at the party is not much fun.
- Listen to every track once from start to finish. A spoken intro, a shout or a choir passage may still be partly audible, and it is better to notice it before your guests do.
- Name the files clearly. Each download is named after the song plus the part, such as “instrumental” or “custom mix”, which keeps a playlist tidy.
- Check the level. An instrumental can feel quieter than the original, because the voice made up part of the loudness. Turn up the speaker rather than pushing every stem to 150%, which can make loud passages clip.
- Keep the original songs at hand in case someone wants to sing along with the singer.
Karaoke at home, on stage and online
Making a karaoke track for your own living room, singing practice or a lesson is fine. As soon as the track leaves your private circle, the rights of the songwriters and the record label come into play. Public performance, copies and derivative versions are among the exclusive rights the U.S. Copyright Office lists for copyright owners.
For a public event, ask the organizer how the music is licensed. Uploading a karaoke video or a cover made over the instrumental needs permission for the recording and the song. This is general information, not legal advice.
Questions you might still have
Is this a free karaoke maker?
Yes. Removing the vocals turns any song into a karaoke backing track, free, without an account, daily limit or watermark. Ads appear around the tool on the page, never in the files.
Can I make a karaoke version of a song from a video?
Yes, if the video is a file on your device. The tool reads the audio track of MP4, MOV, WebM and MKV videos and gives you WAV files. It does not download anything from video sites.
Can the tool change the key of the song?
No. It separates and mixes the stems. Change the key afterwards in a music app, karaoke app or audio editor that has a pitch control.
Does my karaoke app accept WAV files?
Most players and many karaoke apps play WAV. If an app needs MP3, convert the WAV in an audio program you trust. Keep the WAV as your master copy.
Why do I still hear the choir in the chorus?
Backing choirs that blend with the band are harder to separate. Some of them can land in other instruments. A quiet guide voice usually hides this better than any setting.
Before you go
Make two versions of each song, one clean and one with a quiet guide voice, and listen to both once before the evening. That small step solves most karaoke surprises before anyone picks up the microphone.
Sources & useful details
- Meta Research: Demucs music source separation (GitHub)
- ONNX Runtime: how to add machine learning to your web application
- MDN Web Docs: WebGPU API
- U.S. Copyright Office: What is copyright?
- U.S. Copyright Office: Fair use index
These guides are written from the current LoschaZ tools and the sources listed above. Examples are illustrations, not tests on every device or promises of an identical result.
Vocal Remover
Open the free tool
Free · no account
Open the free tool


