A guitarist you play with sends a rough phone recording of a new song. The vocal sits somewhere in the middle, the drums are loud, and you want a version you can practice over before the next rehearsal. You open ShengXi Vocal Remover, separate the track, and then face a quieter question: which version deserves to be kept, and how do you save vocal separation results without collecting a pile of half-finished attempts? This walkthrough follows one session from import to saved result, with the checkpoints that decide what stays.

Before touching the app: define what "good enough" means

The session works best when the goal is narrow. For this example, suppose the goal is a practice backing track where the vocal is low enough to sing against but still audible as a guide. That is different from a clean instrumental for performance, and it changes what you keep.

  • Practice backing: accompaniment forward, vocal faint or removed, timing intact.
  • Vocal reference: vocal kept for learning phrasing, accompaniment turned down.
  • Comparison only: a quick check of whether the arrangement works, not a file to archive.

Write the goal down in one sentence before importing. Every later decision about saving comes back to it.

Step 1: One import, two processing passes

Import the recording from Photos or from a local audio file. The extraction and separation happen on the device, so the source stays where it is. Start with the fast processing mode to hear whether the separation direction is right at all.

Checkpoint: does the accompaniment keep its rhythm and does the vocal land in roughly the right place? If yes, move to fine mode. If not, the source may be too noisy or too compressed to rescue, and no amount of reprocessing will fix that.

Run the same file through fine processing next. This is not about finding a winner in the abstract; it is about hearing which pass suits your stated goal. A fine pass that preserves more detail in the accompaniment may be the keeper for practice, while the fast pass is enough for a quick comparison.

Step 2: Judge with the volume controls, not by memory

Use the mixed preview and the separate vocal and accompaniment volume controls to audition the result. Move the vocal down until you can sing over it, then bring it back up slightly to confirm the melody is still findable. Do the same in reverse if your goal was a vocal reference.

  1. Set the vocal low and the accompaniment high. Listen for rhythmic artifacts or dropouts.
  2. Set the vocal high and the accompaniment low. Listen for bleed from the backing into the vocal.
  3. Return to your goal setting and listen once more without adjusting anything.

That last unadjusted listen is the decision point. If the balance you chose still holds up, this is the version to keep.

Step 3: Name the keeper before saving

Save vocal separation results only after you can describe them. A simple label works: song title, processing mode, and purpose. For example, "New song — fine mode — practice backing" or "New song — fast mode — timing check." The label matters more than the file count, because recent-history storage is only useful if you can tell entries apart later.

If two versions both seem usable, keep the one that matches the goal and discard the other. Keeping both usually means you have not decided yet.

Step 4: Export or share while the context is fresh

Export the keeper, or share it directly to the person who sent the recording. Sharing at this point has a practical benefit: the guitarist can confirm the timing and key feel right, which tells you whether the saved version is actually the one to rehearse with. Do this before cleaning up, not after.

Remember that permission matters. The app does not grant rights to music or video, so confirm you are allowed to process and share the recording before sending it anywhere.

Step 5: Clean up the session, keep the history

Once the export is confirmed, use cleanup to remove the attempts that did not survive the checkpoints. Recent-history storage is there for the versions you might revisit, not as an archive of every pass you tried. A short history with clear labels is easier to work from next week than a crowded one.

What this session teaches about saving

The session above is an example, and your own recordings will behave differently. Results vary with source quality, mixing, noise, and compression, so a setting that worked here may not transfer to the next file. What does transfer is the sequence:

  • State the goal before importing.
  • Use fast mode to check direction, fine mode to judge detail.
  • Audition with the volume controls instead of relying on memory.
  • Label the keeper, share it for confirmation, then clean up.

Save vocal separation results at the moment you can name them, not at the moment they appear on screen. That single habit keeps a practice library small, legible, and ready for the next rehearsal.