How to Make a Karaoke Track from Any Song

Karaoke tracks are simply songs with the lead vocals removed, leaving only the music. For years, you had to rely on whatever karaoke versions were commercially available—or search for fan-made tracks of varying quality. With AI-powered audio separation, you can now create an instrumental version of almost any song yourself.

Here’s how to do it.

What You’ll Need

• A DRM-free audio file of the song (MP3, WAV, FLAC, M4A, or AIFF)
• SongSplit AI for Mac or iPhone

A note on DRM: Songs purchased as downloads from iTunes, Bandcamp, Amazon Music, or similar stores are DRM-free. Songs streamed or downloaded through a subscription (Spotify, Apple Music streaming tier) are protected and cannot be processed.

Step 1: Get the Song File

If you already have an MP3 or FLAC, you’re ready. If you purchased the song from the iTunes Store, locate it in your Music app library and make sure you’ve downloaded it (not just added it to your library).

Step 2: Open SongSplit AI and Import

Drag and drop your audio file onto SongSplit AI, or use the file picker. The app will load the waveform immediately—no waiting for uploads.

Step 3: Choose Your Quality Mode

SongSplit AI offers two processing modes:

• Fast mode — Ideal for a quick preview or casual karaoke use. Processes in seconds.
• Quality mode — Produces a cleaner separation with fewer artifacts. Takes longer but worth it for performances or recording.

For karaoke, quality mode is generally the better choice if you have a few minutes to spare.

Step 4: Process and Preview

Hit process and watch the waveform split into two color-coded tracks—vocals (orange) and instrumentals (green). Once complete, you can preview both tracks directly in the app before exporting.

Toggle between the instrumental and vocal tracks to hear what you’ll get. The instrumental is your karaoke track.

Step 5: Export

Export the instrumental track as an M4A file. You can save it to Files, your Desktop, or anywhere else you’d use it—ready to play from your phone, stream to a speaker, or import into a DAW.

Tips for the Best Results

Start with a high-quality source file. The separation algorithm works on the audio data it receives. A low-bitrate MP3 will produce worse results than a lossless FLAC or high-bitrate M4A.

Some songs separate better than others. Tracks with clearly distinct vocals and instrumentation (pop, rock, country) tend to produce cleaner results. Songs with heavily reverbed or layered vocals—or where instruments and vocals occupy the same frequency range—are harder to separate cleanly.

Expect some bleed-through. AI source separation is impressive technology, but it can’t perfectly unmix a track that professional engineers spent hours mixing. You may hear faint vocal presence in the instrumental. For most karaoke purposes, this is acceptable—the original vocal is low enough that you can sing over it without distraction.

Use headphones to monitor. When checking your exported track, headphones make it easier to hear any remaining vocal artifacts than speakers do.

On Your iPhone

SongSplit AI works the same way on iPhone and iPad. Import from the Files app, process on-device, and export. The whole process happens locally—your audio never leaves your device.

Whether you’re prepping for a karaoke night, wanting to sing along to a song that has no commercial karaoke version, or just exploring what’s possible with AI audio tools, making your own karaoke tracks is easier than it’s ever been.