Start with two files that line up
A mix of a voice and a beat starts from two files: the beat on its own and the vocal on its own. Everything later is easier if they are made the same way.
- Record with headphones. If the beat plays out of a speaker while you sing, the microphone records it too, and that second, delayed copy of the beat ends up inside your vocal file, where nothing can take it out cleanly.
- Export the vocal from the very start of the beat. Silence before the first word is fine, and it is what keeps the two in time. A vocal file that starts at your first word starts at the beginning of the beat when the two are put together.
- Keep the beat out of the vocal file, and effects you are unsure of too. A dry vocal can be given reverb later; reverb that is already printed cannot be taken back out.
- Use the best quality you have. WAV if your app offers it. A beat bought as a WAV file sounds better at the end than the same beat saved from a video site.
- Name them for what they are, such as “vocal” and “beat”. Mixing software and engineers both read names first.
If the music you want to sing over is a finished song with its own singer, take the original vocal out first with stem separation, and use what is left as your beat.
Set the vocal level first
The level of the voice against the beat decides more than any effect. Too low and the words disappear; too high and the voice sounds pasted on top of a record it is not part of.
Finished records are a useful guide. We measured 50 professionally mixed songs, comparing the loudness of the vocals with the loudness of everything else, in LU (loudness units, the same size of step as a decibel): the median vocal sat 2.2 LU below the rest of the song, and half of the songs fell between 0.9 and 4.2 LU below. Styles differ, and so do songs, but a vocal ten LU under the music was quieter than in 48 of the 50.
That is where a home take usually starts. A vocal recorded to peak safely below full scale measures far quieter than a beat that was mastered for release, so the first move is almost always to bring the voice up, or the beat down, by a lot. You can check both with a LUFS meter, then finish the job by ear at a low listening volume: you should understand every word without straining, and the voice should still sound like part of the song.
Make the voice sit in the beat
Once the level is close, a few moves make the difference between a voice on a beat and a song.
- Clean up the bottom. A high-pass filter on the vocal removes rumble and handling noise the beat’s bass does not need competing with.
- Even out the words. A compressor keeps the quiet words audible and the loud ones from jumping out. Gentle settings first; a vocal that sounds squashed on its own is too much.
- Tame the s sounds. Compression and brightness make them louder, and a de-esser turns down only those moments.
- Give it a space. A little reverb or delay, on a send rather than on the vocal itself, places the voice in the same room as the music. Less than you think: if you can hear the reverb as an effect, it is probably too much.
How to EQ vocals goes further into tone, which is where most of the clarity comes from.
Leave the beat mostly alone
A beat is already a finished mix, so you cannot rebalance its drums or its bass, and you rarely need to. If the voice still struggles in a busy part, a small dip in the beat’s upper midrange, where the voice carries its words, makes room. The gentlest version dips only while the voice is singing.
What an automatic mix does with them
Euphona’s mixing takes the two files as stems and lines them up from their first sample. Name them so it can tell which is which: a file called “vocal”, “vocals”, “voice” or “acapella” is read as the voice, and “beat”, “instrumental” or “karaoke” as the backing. A name that says neither is listened to, and if that is not conclusive the mix stops and asks you.
Turn on These stems are unmixed when the vocal is a raw take. If the vocals then sit more than 4.2 LU under everything else, lower than in three of every four songs we measured, they are raised together to that level, and the s sounds of the lead vocal are tamed where they stick out. A vocal already louder than that is left where it is. Leave the switch off when you set every level yourself before exporting: the levels you chose are then kept.
Turn on Offer vocal placements to hear the voice 3 dB further forward and 3 dB further back as well, at matched loudness, and keep whichever sounds right for your song. Add a reference track if you want the overall tone moved towards a record you like, and choose one close to your song.
It does not add reverb, change your timing or correct your pitch, and the beat itself is never rebalanced. The result is a mix with room left at the top, ready for mastering.
Questions
Do the two files need to be the same length?
No, but they need the same start. The mix runs as long as the longer file, and both begin at their first sample, so a vocal exported from the start of the beat lines up and one that starts at your first word does not.
Can I add more than one vocal?
Yes. Lead, doubles and harmonies can each be their own file, and when the stems are declared unmixed they move together, so the balance between them stays the one you recorded.
My vocal is out of tune. Will mixing fix it?
No. Mixing sets level and tone. Tune the vocal file first with vocal tuning, then mix the tuned file with the beat.
Keep reading