AI karaoke maker · Demucs v4 on cloud GPUs
Karaoke Maker: Turn Any Song Into a Karaoke Track With AI
Upload an MP3 and this AI karaoke maker, built on the open-source Demucs v4 model, returns two files: your karaoke track and the vocals on their own.
15 free credits every month · No card needed · Uploads auto-deleted
- 13 sto process a 4-minute song
- 3 creditsper karaoke track
- 5 songsfree every month
- 320 kbpsMP3 downloads
Hear the AI Karaoke Maker on a Real Song
A real result from this online karaoke maker: a Creative Commons song split by the Standard model. Solo the instrumental to hear the karaoke track.
- Vocals
- Instrumental
How to Make Karaoke Online in Three Steps
1. Upload Your MP3
MP3, WAV, FLAC, M4A, AAC or OGG. Free accounts take songs up to 6 minutes and 50 MB, paid plans 15 minutes and 200 MB.
2. Keep 2 Stems Selected
Vocals and instrumental, the mode the Demucs project calls karaoke mode. Pick Standard or Best quality and see the credit cost first.
3. Download Your Karaoke Track
Solo the instrumental in the player to check it, then download it as a 320 kbps MP3, with the vocals as a second file.
What You Get From This AI Karaoke Maker
Two files from one upload. The instrumental is your karaoke track; the vocals come with it as a guide.
- 320 kbps MP3, or WAV and FLAC on Studio
- Original key and tempo
- One by one or as a ZIP
The Karaoke Track (Instrumental)
Drums, bass, guitars and keys: the real recording minus the voice, not a MIDI cover, so it sounds like the song you know.
The Vocals, as a Second File
The singing on its own, to learn the melody and phrasing before you sing. After the acapella itself? Our vocal remover covers that side of the split.
Why an AI Karaoke Maker Beats the Old Tricks
Most karaoke tracks are covers or MIDI remakes, and the old do-it-yourself trick guts the mix. Demucs v4, from Meta AI Research, separates the recording itself.
| AI karaoke maker | Centre-channel cancelling | Cover or MIDI karaoke | |
|---|---|---|---|
| Uses the original recording | Re-recorded | ||
| Keeps the kick, snare and bass | Cancelled with the voice | ||
| Leaves no ghost of the vocal | Faint traces on heavy reverb | The reverb stays | |
| Works on any song you have | Centred vocals only | Only songs someone covered |
9.0 dB Demucs v4, fine-tuned
5.9 dB Spleeter, an older separator
Separation quality (SDR) on the MUSDB18-HQ benchmark, higher is cleaner: the gap between a faint singer left in the track and a track ready for the room.
MP3 to Karaoke: Standard or Best Quality?
Both models run this AI karaoke maker on two stems. Figures are for a 4-minute song.
Standard
htdemucs- Time
- 13 s
- Cost
- 3 credits
Best for: Karaoke nights, practice, most songs
Best quality
htdemucs_ft- Time
- under32 s
- Cost
- 5 credits
Best for: Covers you will record or publish
Measured on our cloud GPUs in September 2026: Standard with two stems, Best quality with four, so two finish faster. The first job after a quiet spell can take a minute longer. Compare the two models in htdemucs vs htdemucs_ft .
What People Make With an AI Karaoke Maker
Karaoke Nights at Home
Karaoke tracks of the songs your guests actually request, in the original arrangement, for any speaker or app that plays an MP3.
Singing Practice
Rehearse with the real band. Solo the vocals to learn the phrasing, then switch to the karaoke track and sing it yourself.
Covers and Open Mics
Record a cover over the original backing track, or bring one to an open mic with no band. Best quality keeps it cleanest.
Choirs, Bands and Teachers
Give a choir or a student a practice track without the lead singer, so they hear their own part against the arrangement.
Tips for a Cleaner Karaoke Track
Start from the best copy you have. A WAV, a FLAC or a 320 kbps MP3 gives the AI karaoke maker more to work with than a low-bitrate rip.
Use Best quality when you will record over the track. For a party, Standard is plenty and costs fewer credits.
Check the karaoke track on headphones before the night. Faint vocal traces show up there first, mostly on heavy reverb or pitch correction.
Expect the backing vocals to go too. Every voice lands in the vocals file, so the karaoke track has no harmonies left.
Need the drums or the bass on their own as well? The stem splitter gives you every part of the band.
What This AI Karaoke Maker Does Not Do
It is a karaoke track maker, not a karaoke video maker: you get an audio file for the player, app or DAW you already use.
- Lyrics on screen
- Video
- Key or tempo changes
- YouTube or other links
- Keeping the backing vocals
Karaoke Maker FAQ
Quick answers on how the karaoke maker handles credits, files and rights. Anything else? Get in touch .
Is this AI karaoke maker free?
Yes, to start. Every account gets 15 free credits a month once you sign in with Google or confirm your email, so you can make karaoke online free for about 5 songs a month: a karaoke track from a 4-minute song costs 3 credits on the Standard model. No card is needed, and credit packs need no subscription.
Can I turn an MP3 into a karaoke track?
Yes. MP3 to karaoke is the most common job for this AI karaoke maker: upload the MP3, keep two stems selected, and download the instrumental. WAV, FLAC, M4A, AAC and OGG work the same way, and the karaoke track comes back as a 320 kbps MP3, or as 16-bit WAV or 24-bit FLAC on the Studio plan.
Does the karaoke track keep the backing vocals?
No. Every voice, lead and harmonies alike, comes out in the vocals file, so the karaoke track is fully instrumental. Sing the harmonies yourself or add a second singer.
What files does the karaoke maker give me?
Two 320 kbps MP3 files: Instrumental, which is the karaoke track, and Vocals. Download them one by one or as a ZIP. On the Studio plan you can choose 16-bit WAV or 24-bit FLAC instead.
How long does it take to make a karaoke track?
About 13 seconds of processing for a 4-minute song on the Standard model, plus the upload and any queue. The first job after a quiet spell can take up to a minute longer while the model warms up.
Can the AI karaoke maker change the key or tempo?
No. It keeps the original key and tempo. Most karaoke apps and every DAW can shift the pitch of the MP3 afterwards.
Does it work with songs in any language?
Yes. The model listens for voices, not words, so the AI karaoke maker works on songs in any language. As with any song, the result depends on the mix, not the lyrics.
Can I paste a YouTube link?
No. The AI karaoke maker works only on files you upload, and you need the right to process them.
Can I use a karaoke track at a public event or in a video?
Only upload audio you have the right to process. Making a karaoke track does not give you new rights to the song: playing it in public, publishing a cover or putting it in a video needs the appropriate licence. Our terms of service and acceptable use policy set out the rules.
Make Your First Karaoke Track Free
Upload a song and hear what the Demucs v4 AI karaoke maker can do. 15 free credits every month, no card needed.
Make a Karaoke Track FreeOut of free credits? See pricing for credit packs with no subscription.