You have a forty-minute window before dinner, one song you want to sing, and a recording that includes the original singer. The goal is modest: get vocals out of a music file well enough to rehearse the harmony line without the lead voice competing with yours. This walkthrough follows that single session from start to finish, with checkpoints where you decide whether to keep going, change a setting, or accept what you have.

Checkpoint 0: Decide What "Done" Means Tonight

Before opening the app, write down one sentence about the outcome you want. For this session, it is: "I need a backing track I can sing over at speaking volume, and I want to hear my own pitch against the original vocal once." That sentence matters because it caps your expectations. You are not mastering a release; you are preparing rehearsal material.

Two practical reminders for this stage:

  • Have permission to process the recording. A file you bought, a clip you recorded yourself, or a track you are allowed to rehearse with. The app does not grant any rights to the music or video you bring in.
  • Know your source. A clean CD-quality file usually behaves differently from a phone recording of a room. Results vary with source quality, how the track was mixed, and how heavily it was compressed.

Checkpoint: if you cannot describe the source file and your permission to use it, stop and sort that out first. Everything downstream depends on it.

Step 1: Bring the File In and Listen Once Before Touching Anything

Import the audio file from your device, or import a video from Photos if your song lives inside a video. Extraction, separation, and export all happen on the device, so the file stays with you.

Now play the original once, all the way through, and listen for three things:

  1. Where the lead vocal sits. Front and center, or pushed back behind the band?
  2. What else lives in the same frequency range. Guitars, strings, and synth pads often crowd the vocal, and those overlaps are what make separation messy.
  3. Any noise or distortion. Hiss, clipping, or a phone-mic recording will travel through the separation along with the music.

This listen is not busywork. It tells you which parts of the result you should judge fairly and which you should simply accept.

Checkpoint: if the track is a rough live recording with heavy room noise, expect a rougher split. Decide now that you will rehearse with it rather than reject it.

Step 2: Choose Fast or Fine — and Know Why

The app offers two processing modes: fast and fine. For a first pass in a short session, fast is the sensible start — it gives you a working split quickly so you can hear whether the song is even usable for practice. Switch to fine when you have decided the track is worth the extra wait, for example when the vocal sits deep in the mix or the arrangement is dense.

A reasonable sequence for one evening:

  • Run fast on the full song.
  • Listen to the separated vocal and accompaniment.
  • If the accompaniment still pulls your ear toward the original singer, run fine and compare.

Keep both results in recent history so you can flip between them instead of re-running and guessing.

Checkpoint: compare the two versions on the same short passage — the chorus, usually. If fine does not clearly help that passage, keep the fast result and move on. Time spent re-listening is time not spent singing.

Step 3: Judge the Split With Three Specific Questions

Play the accompaniment alone and ask:

  1. Can I still hear the original singer? Some bleed is normal, especially on dense mixes and compressed sources. The question is whether it distracts you while singing.
  2. What disappeared that I need? Sometimes a guitar or keyboard part sits in the vocal range and gets pulled into the vocal track. Note it, but do not assume you can fix it with a slider.
  3. Does the accompaniment sound hollow or watery? That is a common artifact when the vocal is lifted out. It is a listening decision, not a failure.

Then play the vocal alone. You are listening for whether it is clean enough to use as a reference — not whether it is perfect. A slightly thin vocal track is still useful for checking your pitch and timing against the original phrasing.

Checkpoint: if the vocal-alone track is too rough to follow, do not force it. Use the accompaniment for rehearsal and skip the comparison step.

Step 4: Set Your Practice Mix and Rehearse

This is where the session becomes useful. Switch to the mixed preview and set the separate vocal and accompaniment volume controls to match how you actually practice:

  • Backing loud, vocal low. The standard rehearsal setting: enough original voice to remind you of phrasing, not enough to lead you.
  • Backing loud, vocal muted. Your performance take — sing over the accompaniment alone.
  • Both moderate. A balance pass for checking whether your line sits against the harmony.

Record nothing, judge nothing, and sing the chorus three times at each setting. Notice which balance lets you hear yourself without straining.

Checkpoint: if you keep raising the accompaniment to cover bleed from the original vocal, that is a sign to try the fine result rather than keep twisting the slider.

Step 5: Close the Session Cleanly

Before you leave, do three small things:

  1. Save or share the version you will use next time. Export the accompaniment mix you settled on so tomorrow's session starts where this one ended.
  2. Note one adjustment to try. For example: "Run fine mode on the bridge, where the original voice is loudest."
  3. Clear what you do not need. Old attempts pile up in recent history; cleanup keeps the next session quick.

Then stop. The value of an evening like this is not a flawless split — it is a usable practice track and a clear idea of what to adjust next time. Separation quality depends on the file you brought in, and no mode will turn a noisy, heavily compressed recording into a studio stem. What you can control is the decision-making: one listen before processing, a deliberate choice between fast and fine, three honest questions about the result, and a mix that serves your voice rather than the original singer's.

If tomorrow's session starts with the same file and a better question, tonight did its job.