How to Isolate Vocals From a Song and Get a Clean Voice Track

To isolate vocals from a song, you run the finished track through an AI separation model that estimates the voice and writes it out as its own file. The Acapella Extractor does this in a browser: upload the song, press one key, and download the vocals as an MP3. Below is what you get, how to judge it, and what to do when it is not quite clean.

What you are actually getting

A finished song is one flattened recording. The vocal, drums, guitars and effects were mixed together, and there is no switch that unmixes them. Separation software looks at the sound and predicts which parts are voice. What you download is a best estimate, not the studio vocal take.

That is useful to know before you start, because it sets the right expectations. A good separation will sound like a clean voice with the music removed. A weaker one will have small pieces of the band in the gaps between words, or a slightly watery quality on long notes.

Why you might want the voice alone

• Remixing and mashups. You need a voice that can sit on a new beat.
• Sampling. A single phrase or word, cut out and played on a keyboard or sampler.
• Learning to sing the song. Without the band, you hear the breaths, the timing and where the singer slides between notes.
• Understanding lyrics. Words are easier to catch without a loud mix over them.
• Studying vocal technique. Teachers and students play a voice alone to hear how it was done.
• Video and podcasts. Putting a voice over new footage or a different backing track.

Step by step with the Acapella Extractor

1. Choose the best file you have. Lossless files and high-bitrate MP3s carry more detail. The tool takes MP3, WAV, M4A and FLAC, up to 10 minutes long.
2. Press "Extract the vocals". The key shows the price before it runs. Sign in with Google when asked. A normal song takes a few minutes at most, and longer when many people are in the queue.
3. Play the Vocals reel first. Click the reel to hear the voice alone, with a small waveform beside it that shows where there is sound. Listen on headphones. The Instrumental reel holds the rest of the song, saved as a second file.
4. Download the Vocals file. It comes as an MP3 at 320 kbps.

Cost is 1 credit per minute of song, and a new account gets 5 credits, which covers one song up to 5 minutes long. If the job fails, the credits are returned.

What to listen for

Press play and scrub through the file, stopping at these places.

• The gaps between phrases. Leftover drum hits, hi-hats and guitar noise show up here first.
• The ends of long notes. Reverb tails are the most common give-away, and a few may carry bits of the instruments behind them.
• The loud choruses. This is where the model has the most to separate, so it is a fair test of the file.
• Doubled or stacked vocals. Several similar voices at once can sound slightly phasey after separation.

If a leftover bothers you, do not throw the file away yet. Sometimes it vanishes once you add your own music underneath.

Which songs separate best

Songs with a dry, upfront voice, which is common in pop, R&B and hip-hop, tend to give the cleanest results. Older recordings made in one room with the whole band playing together, live albums, and songs with a wall of guitars give you more leftovers. The model is strong, but it is separating a mixed recording, and the physical mix sets the limit.

Cleaning up the vocal afterward

• Reverb. If the voice has a long echo baked in, run the file through the Reverb Remover. It splits the vocal into a drier version and the reverb on its own.
• Gaps. An editor's noise gate or a manual fade in the silent parts can hide small leftovers.
• Key and tempo. Change them after you download, on the vocal file alone.

Turning the voice into notes

A clean voice makes a good source for a melody. Upload the vocals file to the audio to MIDI converter, which is a different tool that runs in your browser, and it will give you the notes as a MIDI file you can drop into a music program. A whole mixed song would give a confusing result, but a single voice converts much more clearly.

Keeping the lead and dropping the choir

All vocals stay in the extracted file, backing vocals and ad-libs included. If you only want the lead singer, use Separate Lead and Backing Vocals. Then you can pick which one to keep.

A note on rights

You are working with someone else's recording and a vocal performance that belongs to the artist and label. For private practice and study that is fine. For a remix or sample you plan to release, you need permission from the rights holder.

FAQ

Can I isolate vocals from any song?

You can upload any MP3, WAV, M4A or FLAC file up to 10 minutes long. How clean the vocal is depends on the mix. Dry studio recordings do well, and live recordings do worse.

What is the difference between isolating and removing vocals?

Isolating keeps the voice and discards the music, and removing vocals does the opposite. Both come from the same separation, so each run gives you both files.

Is the isolated vocal the same as the studio acapella?

No. A studio acapella is recorded and mixed on its own. An extracted one is estimated from the finished song, so it can carry small traces of the instruments.

Can I extract vocals from a video?

You need an audio file. Save the soundtrack of the video as an MP3 or M4A, then upload it.

How do I separate vocals from a song on my phone?

Open the Acapella Extractor in your phone's browser, upload the song, and download the vocals when the job finishes. The processing happens on our side, so your phone only handles the upload.