You recorded a song off a video you already had — maybe a live clip a friend sent, or a file saved from a past project — and now you want the music without the singing on top. Before you commit an evening to it, it helps to know what kind of source you are working with, because that single detail shapes the result more than any setting. This article walks through the questions that actually decide the outcome, in the order you will meet them.
First question: is your source good enough to separate?
Separation does not recover something that was never clearly there. The app analyzes a mixed track and estimates which parts belong to the voice and which belong to the accompaniment. When the vocal sits well forward in the mix, that estimate tends to hold together. When the voice is buried, doubled by harmonies, or competing with a dense arrangement, the line between the two gets blurrier.
Compression and noise matter too. A video that has been re-encoded several times, or recorded on a phone at a distance, carries smeared detail that makes the separation less clean. Expect variation rather than a fixed quality level:
- Clean, centered lead vocal: usually separates most predictably.
- Heavy reverb or layered harmonies: some vocal character may remain in the music.
- Loud room noise or clipping: artifacts can appear in both outputs.
None of this means a rough source is unusable. It means you should listen to the result before deciding how much polish you need.
Second question: video or audio file, and does it change the plan?
ShengXi Vocal Remover accepts videos imported from Photos as well as local audio files. If your song lives inside a video, the app extracts the audio as part of the process, so you do not need a separate conversion step. If you already have an audio file, you can go straight to separation.
One thing worth confirming early: whether the version you have is the one you actually want. A live recording and a studio track of the same song will behave differently, and no amount of processing turns one into the other. Pick the source that matches your purpose — a rehearsal backing track usually wants the steadier, more familiar version.
Third question: fast mode or fine mode?
The app offers two processing modes. Fast mode gets you a result quickly, which suits a first listen or a quick check of whether the source is workable at all. Fine mode takes longer and is the better choice when you intend to keep the output.
A sensible order is to run fast mode first, judge the source, then re-run in fine mode once you are satisfied. This avoids spending time on a file that was never going to suit your purpose.
Treat the first separation as a test of the source, not a final verdict on the song.
Fourth question: how do you judge the result before exporting?
Use the mixed preview to hear vocals and accompaniment together, then switch to the separate tracks and adjust their volume independently. This is where you find out whether the separation suits your goal. Listen for three things:
- Vocal bleed in the accompaniment. Faint traces of the singing may survive. On a loudspeaker this is often inaudible; on headphones it can be obvious.
- Instrument loss in the vocal track. Some instruments sitting in the vocal's frequency range may travel with it.
- Balance after adjustment. Lowering or raising one side can hide small imperfections or expose them, depending on the material.
Adjust the two volume controls until the balance works for your use. If you are preparing a backing track to sing over, a slightly reduced vocal presence in the mix can act as a guide without dominating.
Fifth question: what should you keep, and what should you delete?
Recent-history storage means earlier separations stay available if you want to return to them. That is useful when you want to compare a fast-mode result with a fine-mode one, or revisit a file after changing your mind about the balance.
It is equally worth cleaning up files you no longer need, especially if you are working with several versions of the same song. Sharing is available for results you decide to keep.
Before you publish or perform with it
Processing audio on your own device does not change who owns the song. You need permission to process and publish your source material, and the app does not grant any rights to the music or video. Practicing privately with a backing track and posting it publicly are different situations, so check where your use falls before sharing anything.
The short version: start with a source whose vocal is clear, test it in fast mode, switch to fine mode when you intend to keep the result, and judge the output through the separate tracks rather than the mixed preview alone. Do those four things and you will know quickly whether the instrumental you get is the one you can use.