How to remove vocals from a song in 5 steps
- Open the Vocal IsolatorIt is already open at the top of this page, inside the Music Studio.
- Add your songChoose Upload a file (MP3, WAV, M4A and other audio up to about 9 minutes, or an MP4 or MOV video up to 60 MB), or Paste a link to an audio or video file.
- Under the form.
- Press Isolate VocalsSpeechma Pro’s AI splits the song into two parts: the voice and everything else.
- Download what you needSave the music-only track for karaoke or a backing track, and the vocals for remixes, practice or transcription. Both are downloadable.
Which file should you use?
| You want to | Use this file | Tip |
|---|---|---|
| Sing karaoke or practice a song | Music-only | Pair it with the lyrics on screen |
| Rehearse with a band or choir | Music-only | Lower the tempo in an editor if needed |
| Learn harmonies or phrasing | Vocals | Loop one line at a time |
| Write down the lyrics | Vocals | Run it through Speech to Text with sentence timestamps |
| Remix over a new beat | Vocals | Only with songs you have permission to remix |
| Hear a new singer on the song | Either | Try AI Voice Cover for a new voice style |
How to make a karaoke track
The music-only track is your karaoke track. To turn it into a full sing-along video:
- Get the instrumental: download the music-only file from the Vocal Isolator.
- Time the lyrics: upload the vocals file to Speech to Text with timestamps on and download the SRT. You now have every line with its start and end time.
- Build the video: in CapCut or any editor that imports SRT captions, put the instrumental on the audio track and import the SRT as captions. Correct any misheard words.
- Keep a guide vocal (optional): mix the vocals back in very quietly to help singers who do not know the song well.
For more ideas, see the karaoke track maker page.
Honest limits: what about vocal bleed?
AI separation works from a finished stereo mix, not the original studio tracks. It is very good, but not magic.
- Some bleed can remain on dense mixes. Loud rock, heavy distortion, big reverb and stacked harmonies can leave faint traces of music in the vocals, or a ghost of the voice in the music-only track.
- Backing vocals may stay in the music. Choirs, ad-libs and harmonies mixed like instruments can be treated as part of the band.
- Effects travel with the voice. Echo and reverb on the singer may be split between both files.
- Source quality matters. A WAV or a high-bitrate MP3 separates far better than a low-quality rip or a phone recording of speakers.
- Live recordings are harder. Crowd noise and room echo blend into the voice.
Tips for the cleanest result
- Use the best file you have. Avoid recording a song off speakers with your phone.
- Trim to the part you need before uploading, for example just the chorus for a Reel.
- Listen on headphones to the quietest and busiest parts of the song.
- Hide small leftovers with a gentle noise gate or a slight EQ cut in an editor such as Audacity or GarageBand.
- Longer than 9 minutes? Split it into parts, process each one, and name them in order.
Copyright: practice is fine, sharing needs permission
Removing vocals for private practice, a home karaoke night or study is one thing. Publishing an instrumental or a vocal track from someone else’s song is another: the song’s composition and recording still belong to their owners. Before you upload a karaoke video, a remix or a cover to YouTube, TikTok or Spotify, make sure you own the song or have permission, or use the platform’s licensed music tools. Read our Copyright & DMCA policy.
Want music you can publish as you like? Make your own with the AI Song Generator, which includes commercial use.
Frequently asked questions
How do I remove vocals from a song?
Can I get the instrumental and the vocals separately?
How long can the song be?
Why can I still hear a little singing in the instrumental?
Can I remove vocals from a YouTube video?
Is it legal to post a karaoke version of a song?
Make better voiceovers, faster
- New voices and features the day they launch
- Script templates and tricks that make AI voices sound human
- Creator stories, community picks and early access