Work Out Which Audio Problem You Have, Then Pick the Tool
You film something good. You drop a song under it. You post it. Twelve minutes later the audio is gone, or the whole thing is muted, or it just quietly stops showing up in anyone’s feed.
Then you go looking for a fix, and the internet hands you two of them without explaining that they solve completely different problems. One is a vocal remover. The other writes you music from scratch. Worse, one of them does not solve the problem you probably have.
Let us do the unglamorous part first.
Before anything else: stripping the singer out does not make a song yours
This is the part people skip, and it is the one that actually costs them.
Lifting the voice off a track leaves ownership exactly where it was. What sits underneath is still the same copyrighted recording. Platform detection matches against the master audio, and the backing is most of that master. Putting a chart song through a vocal remover, then expecting the claim not to fire, is optimistic in a way that gets people muted.
So if your goal is “post a popular song and not get caught,” neither tool below is going to save you. Stop reading, or read on and change the goal.
Still here? Good, because the two legitimate problems are worth solving properly.
Problem one: the singer keeps competing with you
This is the real use for separation, and it has nothing to do with dodging anything.
Say you are talking over the clip, or singing over it, or the lyrics keep pulling attention away from what is on screen. Hand the track to a vocal remover. What comes back is the singing as one file and the accompaniment as another. Put the accompaniment under your own audio and nothing is crowding you out of the same frequencies any more.
Where a vocal remover genuinely applies: practice recordings, karaoke at home, a cover you are working on, a rehearsal file, anything private, or anything where the rights are already sorted. It is a production tool, not a workaround.
The result depends heavily on the song. Dense mixes with layered harmonies fight it. A track drowning in reverb comes back thinner than expected, because a vocal remover chases those tails and takes some of the room with them. Judge it before you build anything around it.
Problem two: you want music nobody can claim
Different problem, different tool, and this is the one most people should have reached for.
If what you actually need is something under your clip that will not get flagged, the answer is not to modify someone else’s song. It is to have music that was never theirs. You can generate original music from a written description: say the genre, the mood, roughly the tempo, what it is for, and get tracks back built to that brief.
Two details that matter in practice. Each run costs 20 credits and hands back two takes on the one brief rather than a single track, so comparison is part of the process instead of a second attempt. And there is an instrumental-only toggle, which is usually what you want under a talking clip, because a generated vocal singing over your own voice recreates the problem you just escaped.
Before your next post: which one is it?
Three questions, in order:
Is anyone else going to see this? Private practice, a file for your own use, something in a group chat — a vocal remover is fine and simple here. Public, monetised, or on a platform with content matching — the source track’s owner is still in the picture no matter what you did to the file.
Are you speaking or singing over it? If yes, you need the vocal out of the way, or you need something written to sit behind a voice. Both routes work; the rights question decides which.
Does it need to sound like that specific song? If yes, you are committed to the licensing conversation. If no — and honestly, usually no — generated music sidesteps the entire issue, and you get to specify exactly the mood you wanted anyway.
Where people go wrong
Four patterns, all avoidable.
Testing on the intro. Almost every song opens sparse, and a vocal remover handles those opening seconds beautifully. The chorus is where things fall apart. Check the busiest part or you have learned nothing.
Judging on phone speakers. Artefacts hide at low volume and on small drivers. Use headphones, at roughly the loudness your viewers will have.
Assuming a clean result equals permission. Covered above, but it bears repeating, because it is the mistake with an actual cost attached.
Generating music as an afterthought. A vague prompt gives vague output. “Sad piano” is not a brief. What is the clip, how long, what should someone feel at the end of it — that level of detail is what separates a usable track from generic background.
When to skip both
Sometimes the honest answer is that neither tool belongs in the project.
Live recordings with crowd noise defeat a vocal remover, because the voice and the venue arrived together and leave together. Choral or heavy-harmony material has no single lead line to isolate. And if the clip genuinely needs that song — a specific cultural reference, a joke that only works with the original — then no amount of processing gets you around it, and the real task is finding out what a licence costs or picking a different idea.
The tools are good. They are just answers to specific questions, and it helps to know which question you were asking before you start uploading things.
