This month we take a look at ways of tackling vocal pitching and timing problems.
The song 'Time For You' was sent to us by Michael Bowkunowicz because he didn't feel very happy about his vocal performance, and although he'd spent some time trying to fix it up, using Melodyne, he still felt it could sound better. Michael confessed that he was inexperienced at mixing and I think he wanted to see what we'd do with the song — a whimsical little ditty, set against a backdrop of acoustic guitar, bass, keyboard pads and synth arpeggios, as well as a little light percussion and a layered kick drum playing on the on-beat. There were also two tracks of backing vocals, and though I suspect Michael intended using only the better of the two, I thought I'd be able to use both, with a little tweaking.
On loading the track into Logic, I found that Michael had included his 'Melodyned' version of the vocals, as well as an unprocessed one. I decided, after listening to both, that I should go back to the original and see what could be done with it, as there were some very evident pitching problems. Michael occasionally pitches notes either sharp or flat and then sustains them pretty accurately, a problem that can sometimes be due to the way the headphone mix is set up. It's also clear from the performance that he lacks a bit of confidence when singing, so some of the remaining notes either scoop in too much, droop away at the end, or waver a little too much. I initially tried Auto-Tune's Auto mode, set to the song's key of D-major, but Auto-Tune tried to push some of the slurred notes to the wrong pitch, as that was closer to what Michael had sung. As I'm more familiar with the visual side of Melodyne than Auto-Tune, I decided to give that a go instead.
Applying a general pitch correction got most of the notes onto the right centres, but a few were pushed to the wrong pitch because they were originally closer to the adjacent semitone than the intended note. These I dragged back to pitch manually, then I used the modulation and 'pitch droop' tools to reduce the wavering and scooping but without getting rid of the problem entirely. This took about half an hour, and though the result still wasn't perfect it left the vocal performance sounding much more confident, and definitely more musical.
The harmony vocals were easier to deal with and responded well to Auto-Tune's Auto mode, again with a D-major scale dialled in. However, the phrasing on some of the backing parts didn't quite match that of the original, so I sliced up the backing vocal tracks, into separate words where necessary, and slid these along to better match the lead vocal. To give the vocal a more solid sound, I used the Tube-Tech CB1 compressor plug-in to level it out and then applied the Noveltech Vocal Enhancer plug in to give the vocal more breath and edge. I compared this with simple high-end EQ and found that the Vocal Enhancer produced a richer, more credible sound, so I stuck with that. I set up a vocal plate reverb, courtesy of the TC ClassicVerb plug-in (set up as a post-fade send), to 'wet' the lead and backing vocals slightly, and also to treat some of the other instruments where the original seemed overly dry.
Michael had recorded the acoustic guitar parts in sections, and I could also hear some edits in what he'd done, but there was little I could do about those, and in any case they were far less obvious in the context of the whole mix. All I did here was add some compression to thicken up the main guitar sound, using a longish attack time (20-30ms) to allow the initial transient to come through more strongly, then rolled off some low-end and added some 'air EQ' by applying a broad boost at around 10kHz.
During the intro, middle and end sections there's another acoustic guitar part recorded over two tracks that was noticeably less bright-sounding, so I added a touch more air EQ to that. There is also a section in the middle of the song where there's a bit of a break which leaves a decaying acoustic guitar chord exposed, and there were some audible noises here — possibly from finger nails tapping the body or neck. The only way I could reduce their impact was to automate some top cut (low-pass) to lose the top end of that chord as it decayed, then open the filter up in time for the guitar to start up again. Though this didn't eliminate the noises entirely, it did make them less intrusive. It is very important when playing exposed chords to let them ring cleanly: many an otherwise good ending has been spoiled by finger noises, chair squeaks or coughs.
The bass part (a synth) needed only slight EQ to thicken it up a little. I then added some slow flanging and very heavy reverb to the string-like pad part, to add a bit more movement, and to help it fit in with what I felt was the retro vibe of the song. I also added a Roland Dimension D chorus plug-in to a choral synth pad, to give it a bit more width, as well as dialling in a bit of that plate reverb to create a better sense of space.
All the drum and percussion parts were left pretty much as they came, other than for level balancing, and the same is true of the synth arpeggios — though I did sneak a little reverb onto the arpeggios and reverse cymbal hits. Michael had used a tambourine where you'd normally expect a snare and there were no toms, so the whole vibe was more percussion than drums.
I felt that the pronounced kick-drum on the original gave the rhythm a rather awkward, lurching, 'Chas and Dave' feel, so I dropped it in level in my version of the mix, and pulled down the synth arpeggios too, as they tended to be a bit irritating if left too loud. There was also a section near the end where Michael had tried to create a synth-pad harmony to go with his vocals, but one note of this new part clashed awfully with the vocal melody — it sounded like a semitone difference, producing an alarming discord, so I simply copied and pasted earlier incidences of this part for the two pad tracks in place of the offending ones, and this sounded much better to my ears.
Michael Bowkunowicz: "Paul's mix is a considerable improvement, especially considering he had his work cut out, given the relatively poor recording he had to start with!
"The separation is much better — before it was almost non-existent. The dynamics have lifted the song, and it now seems to keep some momentum, whereas before it just sort of trudged along. And what I was looking for but didn't achieve myself was a little more impetus during the 'chorus'; because the verse and chorus use the same chords and phrasing I needed some more contrast, and Paul's achieved that.
"On reflection, I think I do need to re-record the song, and use some different chords in the chorus... and play the guitar better... and sing in tune. Paul suggested that a few singing lessons may help and I'm seriously thinking about doing that. But this was my first attempt at a mix and the results Paul got have really encouraged me, and shown me how better treatment can help lift a song if it has any potential at all."
I used both backing vocal lines, one panned to each side, as they sounded pretty sweet once 'Auto-Tuned'. The two keyboard pads and two arpeggio parts were also panned left and right, as were some of the percussion sounds, but the main vocal, kick drums and acoustic guitars were kept nominally central, as was the bass.
Before setting up the mix balance, I also trimmed out any unused sections of audio, to make sure there were no unwanted noises and to lighten the CPU load — there's no point playing back long stretches of silence from your hard drive, after all. I had to fade some tracks right at the end of the song to get it to end neatly, but, on the whole, balancing the levels was the most important issue after improving the vocal pitching, and the levels of several parts had to be automated. In particular, there is a piano part that only occurs in the middle section of the song, which started well but then got quite messy and uncertain. In the end, I dropped it in level by a few dBs when the vocals were present and this made things sound much tighter. I then gave it quite a lot of reverb to soften the effect.
Because the song starts with just an acoustic guitar and a pad synth behind the voice, I automated the vocal level to bring it up by a couple of dBs when all the instruments and rhythm section joined in — the voice tended to sound over-loud on the intro if left at the same level all the way through the song. If, when you've done all the level automation, you realise the whole track needs to be a bit louder or quieter, you can easily make adjustments using the level control of any inserted plug-in that has one, though you need to make sure this comes after any dynamics plug-ins, or you may need to adjust their threshold setting.
To add density to the whole mix, I used the SSL Bus Compressor plug-in, set to squash by four or five dB, and followed this with the trusty Waves L1 limiter, which I set just to prevent clipping.
In my version of the mix, the vocals took a more dominant position, and the whole mix seemed a touch drier and brighter than Michael's own. Dropping the levels of the kick drum and synth arpeggios smoothed out the rhythm of the song and also made the synth parts sound less dated. I'm not sure quite how far my mix matches Michael's original vision for the song, but comparing it with his, I feel the vocals now sound far more tuneful and confident.
We've put audio files on the SOS web site Michael's original mix, Paul White's mix, and some before and after files of individual instrument parts. If you want to hear them, go to www.soundonsound.com/sos/nov07/mixrescueaudio.htm.