Make a practice track in three steps
- Import the songOpen Woodshed and add an audio or video file from your phone or computer.
- Choose a splitPick vocals and instrumental for karaoke, four stems for most practice, or six stems to pull out guitar and piano as well.
- Mute your partOn the mixer desk, mute the vocal (or your instrument) and press play. Bring it back quietly as a guide when you need it.
How a modern vocal remover works
Old karaoke tricks cancelled out whatever sat in the middle of the stereo mix. That took the vocal out, but it also took the bass, the kick and the snare with it, and left a hollow, phasey track that was no fun to sing over.
Stem separation works differently. A neural network trained on thousands of multitrack recordings listens to the whole song and rebuilds each part on its own: one track of vocals, one of drums, one of bass and one of everything else. Six-stem models also split out guitar and piano. Because each stem is rebuilt rather than cancelled, the band underneath stays full and punchy.
Woodshed runs these models on GPUs in the cloud, so the result is the same on a laptop, a tablet or a five-year-old phone. Your device only has to play the stems back, which it does with a dedicated audio engine.
Removing more than the vocals
Singers want the voice gone. Everyone else wants their own instrument gone. Once a song is split, any stem can be muted, soloed or turned down, so the same import serves a whole band.
- Guitarists mute the guitar stem and play the parts over the real rhythm section.
- Bassists mute the bass, or solo it first to hear exactly what the record is doing.
- Drummers mute the drums and get a drumless track with the click following the band.
- Singers pull the lead vocal down to a whisper so it is there as a guide but not in the way.
- Keys players use the six-stem split to isolate the piano part.
Getting the cleanest result
Separation quality depends on the recording. Studio tracks with a clear mix separate best. Live recordings with lots of room sound, heavily distorted guitars or vocals drenched in reverb leave more traces behind, a faint echo of the singer or a bit of cymbal in the vocal stem.
Start from the best file you have. A lossless WAV or FLAC beats a low-bitrate MP3, and the original master beats a recording off a phone speaker. If a part still bleeds through, Woodshed Pro has a Hi-Fi mode that uses a heavier model and takes about four times longer, which is worth it for a song you will play for months.
A small amount of bleed rarely matters for practice. With the band playing and you playing over it, a ghost of the original part simply reads as a doubled line.
How to practise with a vocal-free track
Removing the part is only the start. The real gains come from what you do next.
- Learn it with the part in. Listen with the original vocal or instrument up so you know exactly how it should sound.
- Turn the part down, not off. Play along at half volume so you can check yourself against the record.
- Loop the hard line. Set an A-B loop on the phrase that keeps tripping you and drop the speed until it is clean.
- Then take it out completely and play the whole song with the band.
- Record a take over the stems and listen back. You will hear things you cannot hear while playing.
Is it legal to remove vocals from a song?
Splitting a song you have a copy of, for your own practice or teaching, is personal use. Woodshed keeps your stems private to your account and never publishes them. Releasing or performing someone else's recording in public is a separate question, and still needs permission from whoever owns it, separated or not.
