Home / Guides / Practise Singing with Backing Tracks

How to Practise Singing with Backing Tracks

Singing along to a commercial recording trains your ear but tells you nothing about your own voice. Singing over a purpose-built backing track, one where you control the mix, is a fundamentally different exercise. This guide shows you how to build that backing track from any song you already know, then use loops and recording to drill the parts that need work.

Last updated August 2026 · 8 min read

Why a custom backing track beats a karaoke track

Karaoke tracks are produced by session musicians and often differ from the original in tempo, key, or feel. If you are learning a specific recording, studying a phrasing choice, matching a producer's arrangement, or auditioning for a show, that gap matters. A backing track derived from the original recording keeps every nuance intact: the reverb tail on the snare, the specific voicing of the piano chord, the exact tempo drift in the bridge.

AI stem separation lets you build that track in seconds. Upload the song, mute the vocal stem, and what remains is a genuine instrumental created from the same recording you are studying. You can also dial in how much of each instrument you want, more piano if you are matching your pitch to that, less drum if the kick is masking your low notes.

If you are new to stem separation, the guide on how to remove vocals from a song covers the process and its limitations in detail. For songs with complex arrangements, you may also want to read about isolating individual instrument stems before you start mixing.

Step-by-step: building your practice session

1. Upload your song

   Go to JamScore AI and drag in an MP3, WAV, M4A, or OGG file. Files up to 10 minutes are supported, all processing happens in the browser, nothing is stored on a server you do not control. On the free plan you get three uploads per month, so it is worth being deliberate about which songs you pick for a given month.

2. Run stem separation

   Click Separate. The AI separation engine analyses the audio and produces individual tracks for vocals, drums, bass, guitar, piano, and other instruments. Processing typically takes under a minute for a three-to-four minute song. Once it is done, you will see the studio mixer with all six stems loaded.

3. Build your backing mix

   Mute the vocal stem entirely. Now adjust the remaining stems to suit your practice goal. If you are working on pitch, bring the piano up so you have a clear harmonic reference. If you are working on timing, bring the drums up and push everything else back. The per-stem EQ, low, mid, high, lets you shape each instrument further: cutting low-mids on the guitar, for instance, can clear space for your voice in the midrange without losing the guitar's presence.

4. Set an A/B loop over the section you want to drill

   Rather than running the whole song on repeat, bracket the specific section causing trouble. Drag the loop markers to the start and end of the phrase, bar, or note run you need to work on. A chorus hook, a high note approach, a syncopated rhythm, pin it down and let it cycle. Repetition within a focused loop is significantly more efficient than running full-song repetitions.

5. Record your vocal take

   Hit Record. Sing over the looping backing mix. The recording captures your microphone input against the instrumental playback. You can record multiple takes without leaving the page; each is saved locally in the session so you can audition and compare them.

6. Listen back critically

   Play your recording alongside the original vocal stem. Listen specifically for pitch (are you landing flat on sustained notes?), timing (are you rushing the approach to a high note?), and tone (does your vowel shape match the original's resonance?). Do not try to fix everything at once, pick one variable per repetition cycle.

7. Iterate the mix as your practice evolves

   As you get more comfortable with a section, gradually reduce the piano or other pitch reference stems. Practising with strong harmonic support is appropriate at the start; weaning yourself off it as you progress trains your internal pitch sense rather than building a dependency on the crutch.

Caveat, vocal bleed in the backing track. AI separation is imperfect, and some songs, particularly those with lead vocals panned to centre and densely layered with instruments, will leave residual vocal artefacts in the backing stems. If you practise against a track where the original vocal is still audible, even quietly, you risk unconsciously pitch-matching to it rather than finding the pitch independently. Check the backing track before you record. If there is noticeable bleed, try enabling de-reverb on the vocal stem before muting it, this can clean up the tail that often leaks across. Alternatively, accept the limit and use the exercise for timing rather than pitch.

Working with different voice types and song styles

The approach above is broadly applicable, but a few adjustments make it more effective depending on the material.

Songs with heavy production or dense arrangements

Pop and R&B productions often stack multiple synth pads, backing vocals, and effects in the "other" stem. This stem can sound cluttered on its own. Reduce its volume and let the bass and drums carry the low-end foundation instead. In heavily produced songs, the bass line is often the most reliable pitch reference after the piano, especially for singers working in a chest-dominant register.

Songs with sparse or acoustic arrangements

Acoustic folk and singer-songwriter material separates more cleanly because instruments are less blended. However, guitar chords can be ambiguous pitch references, if you are trying to nail a specific note, a piano stem (if the original has one) or bass is more reliable. If neither is present, consider transcribing the guitar stem to notation using the MIDI and sheet music conversion tool, then you can identify the exact chord voicings and know precisely which notes you should be landing on.

Songs with lead vocal harmonies

If the song has prominent harmonies, they may appear in the vocal stem alongside the lead. When you mute the vocal stem, those harmonies disappear too. You can use this deliberately, practise one harmony part by isolating the original stem, then record yourself singing that part and compare. The mixer lets you blend your recording against the original stems to hear how well your harmony sits.

A word on pitch training versus pitch dependency

A common mistake when using backing tracks for practice is making the reference mix too full. If the piano is playing the exact note you need to sing, every passing chord is pre-answered for you. This can produce technically accurate singing without developing genuine pitch independence. A disciplined approach cycles through phases: full support early, then progressively stripping back the harmonic stems until you are singing against drums and bass only, and finally testing yourself in silence before returning to the full mix. The A/B loop makes it straightforward to switch between those mix states mid-session by adjusting stem volumes without stopping playback.

Exporting your work

Once you have a backing mix you want to keep, you can export individual stems as WAV files and combine them in a DAW or audio editor. There is no direct "export the mixed backing track as a single file" button, you will need to export stems separately and mix them down externally. Note that the app exports WAV only; there is no MP3 export option.

If you want to take practice further, for instance, by seeing the melody you are trying to sing as actual sheet music, the companion guide on converting a song to MIDI and sheet music walks through transcribing a vocal or instrument stem into staff notation and numbered notation.

Frequently asked questions