You recorded your band's practice on your phone so you could hear how the new song is coming together. Now you want to hear the arrangement without your own voice on top of it, so you can check the guitar lines and figure out where your entry keeps slipping. This walkthrough follows one session of that kind from start to finish, with a checkpoint after each step and a decision about what to adjust next.
The source in this example is a single practice-room recording: one phone track, voices and instruments sharing the same space, some room noise, a little compression from the phone's own processing. That is a normal starting point, and it is also a useful one to reason about, because results always depend on source quality, mixing, noise, and compression.
Before you separate anything: decide what "good enough" means
Separation is not a magic trick, and you should not expect a perfect studio result from a phone recording. Set a target before you touch any settings:
- For arrangement listening, you mainly need the instruments clear enough to follow the chord changes and the drum pattern.
- For rehearsal, you need the accompaniment steady enough to sing over without the original voice pulling you off pitch or off timing.
- For a rough reference mix, you need the balance to sit comfortably in headphones, not to sound like a commercial release.
If the source is noisy or heavily compressed, some bleed and artifacts are expected. That does not mean the file is unusable, only that you should judge the result against your actual purpose.
Checkpoint: write down in one line what you want from the file. If you cannot name it, you will keep re-processing without a way to decide when to stop.
Step 1: bring the practice recording into the app
Open ShengXi Vocal Remover and import the file. In this example the recording came from the phone's camera roll, so it arrives as a video. You can also import a local audio file if you recorded practice on a separate device and copied it over.
One thing worth knowing at this stage: the extraction, separation, and export all happen on the device. The source file does not need to be uploaded to a server, so you can work through this without waiting on a connection and without sending the practice recording anywhere.
Checkpoint: confirm you are working with the right take. Practice sessions often produce three near-identical files, and separating the wrong one wastes the useful part of your session.
Step 2: choose fast processing and listen before you tweak
The app offers fast and fine processing modes. For a first pass on a practice recording, start with the fast mode. The point of this pass is orientation, not final quality: you want to hear quickly whether the separation is usable at all, and whether the problem you are trying to solve is really a separation problem.
Then listen to the result in the mixed preview before adjusting anything.
- Can you follow the guitar and drums without the lead voice covering them?
- Does the voice sit cleanly in its own layer, or does it pull instruments along with it?
- Does the character of the room survive, or does the result sound thin and processed?
Checkpoint: decide whether the fast pass already meets the goal you wrote down. If it does, stop here. If it does not, note specifically what is wrong before moving on.
Step 3: adjust the balance between voice and accompaniment
The result supports separate vocal and accompaniment volume controls, and you can preview them mixed. This is where most of the useful adjustment happens, and the right move depends on what you heard in the previous step.
- If the voice is still distracting, lower the vocal volume until it becomes a faint guide rather than a competing performance. Keeping a trace of it is often more helpful for timing than removing it entirely.
- If the accompaniment sounds thin, raise it rather than pushing the vocal further down. A quiet, balanced mix is easier to work with than an aggressive one.
- If instruments bleed into the vocal layer, accept it or reduce the vocal level. Bleed is a property of the original recording, not something an adjustment will remove.
A quiet vocal guide is one of the most practical results of this step: enough to keep your place, not enough to lead you.
Checkpoint: play the mix once without watching the controls and once while adjusting. If your hands keep moving, you have not settled on a balance yet.
Step 4: switch to fine mode only when it earns its time
Fine mode is the second option, not the default one. Reach for it when the fast pass clearly misses your goal and the material seems worth the extra pass, for example when the vocal layer is muddy in a way that makes the arrangement hard to study.
Do not run both modes out of habit. Compare the two results against the same checkpoint question, and keep whichever one actually serves the task.
Checkpoint: if the fine pass does not change your answer to the checkpoint question, keep the faster one and move on.
Step 5: export, share, and clean up
When the balance is settled, export the result. The app supports sharing, so you can send the accompaniment to whoever else is rehearsing, or keep it on your own device for the next session. Recent results are stored in a history list, which means yesterday's take is still there if you want to compare.
Then clean up. Practice recordings accumulate quickly, and the cleanup option exists for exactly this reason.
Checkpoint: before you share anything, make sure you have permission to process and publish the source material. The app does not grant rights to music or video; that responsibility stays with you.
What this session actually produced
The outcome here was modest and useful: an accompaniment track clear enough to rehearse against, with a faint vocal guide left in place for timing. It is not a studio separation, and it is not meant to be. Working through the checkpoints above kept the session from turning into an endless series of re-processes, because each step ended with a decision rather than another round of guessing.
If you try the same sequence on your own practice recording, the value is in the order of operations: set the goal, do a fast pass, listen, adjust the balance, escalate only if it earns it, then export and clean up. That routine travels well from one take to the next.