The audio files described below accompany the Steal The Feel feature in SOS December 2011. For the best results, download the WAV files and audition them in your DAW, btu we've also made MP3 versions available.
Here's a clean drum break sample at an original recorded tempo of around 95bpm, which I want to slow down to a target tempo of around 85bpm so that it fits with my programmed beat.
For this audio example I've time-stretched the DrumsTempo01_Raw drum loop using Cockos Reaper's built-in Elastique 2.1 Pro pitch-shifting algorithm, operating in its 'Normal' mode. Notice how the tone of the drum sustains takes on a slightly chorus-like quality. The amplitude envelope of each drum hit has also extended by the processing.
In this example I've transfered the DrumsTempo01_Raw drum loop into Celemony's class-leading pitch/time-manupulation software Melodyne Editor, operating in its Percussive mode. Although the chorus-like shimmer is reduced compared with the DrumsTempo02_Elastique example, the drum-hit envelopes have still been stretched and the naturalness of the sound's room ambience has suffered.
Instead of time-stretching the DrumsTempo01_Raw file for this audio example, I implemented the tempo drop using audio editing techniques, chopping the drum break up into its individual hits and then slowing down the timing of the slices. The gaps which this left between the slices were then covered over by extending the start of each slice, with equal-power crossfades to smooth the slice boundaries. As you can hear, the side-effects of this tempo-matching method are much less audible than those of either of the time-stretching approaches.
Here's a guitar loop covered in unwanted hiss. If you try simply EQ'ing to reduce the hiss, you'll also unacceptably dull the guitar tone, so more advanced processing than this is required.
One technique which can pretty reliably achieve hiss-reduction of up to 6dB or so is high-frequency expansion, by virtue of a multi-band dynamics processor. In this particular example I've applied Universal Audio's UAD Precision Multiband plug-in to the GuitarHiss01_Raw file, expanding the signal above a 4kHz crossover frequency at a low 1.24:1 ratio. The attack and release times were set to their minimum values of 50 microseconds and 20 milliseconds respectively, and the threshold adjusted to give approximately 6dB of gain reduction. Although a little of the instrument's subjective high-frequency sustain is lost, you do get a significant reduction in the subjective noise level in return.
For a greater degre of noise reduction than heard in the GuitarHiss02_HFExpander file, you need to start investigating dedicated audio restoration tools. One affordable option here is Voxengo's Redunoise plug-in, which I've used here to demonstrate the degree of improvement you might expect. Firstly I fed the plug-in with a sample of the noise on its own, so that it could build up a noise profile, and then I adjusted the various spectrally dependent parameters with the GuitarHiss01_Raw file playing back so that I could judge their efficacy by ear. Again, you lose a little of the glossiness of the guitar part, but that's a small price to pay for a massive decrease in the background noise level.
Here's the same guitar loop as in the GuitarHiss01_Raw example, but this time with vinyl noise over it, rather than just broadband circuit hiss. Although this gives it an appealing vintage character, the transient spikes in the vinyl noise are distractively sharp-sounding, so the next few examples showcase some ways of tackling this issue.
Applying 6dB of transient reduction using SPL's Transient Designer plug-in rounds off the edges of the crackle peaks a little. However, there's a limit to how effective full-band processing such as this can be -- indeed you can already hear a certain amount of unmusical gain pumping in the guitar line, and the guitar's own lower-frequency transients are also being ducked undesirably.
Another strategy for reducing vinyl-noise crackle is to apply fast limiting to the high-frequency band, which is what I've tried for this audio example, feeding the GuitarCrackle01_Raw file through Cockos Reaper's built-in ReaXcomp plug-in. I've limited frequencies above about 4.2kHz at a ratio of 100:1 with an instantaneous attack using the most peak-sensitive level detection (the RMS size being zero). Release has been set at 4ms time. Although the gain-reduction meter for the high band is registering peaks of up to 25dB(!), this only dulls the guitar tone very slightly because the gain-reduction is mostly triggering only during the extremely brief crackle peaks.
For this audio example I tackled the vinyl crackle problem in the GuitarCrackle01_Raw file using audio editing, rather than plug-in processing. I did this by manually slicing the audio file whether there was a big vinyl-noise click and then placing a crossfade over it to briefly to duck its level -- a crossfade made up of two concave slopes can be used to achieve this. Although this approach doesn't reduce the overall subjective level of the vinyl noise much, it does help make it significantly less distracting in context, and without any detrimental side-effects to the guitar sound whatsoever.
Here's the same guitar sample again, but this time with simulated earth-loop hum/buzz of a type you might encounter in the USA (in other words, based around the mains alternating current frequency of 60Hz).
Although it's best to try to avoid earth-loop hum during the recording process itself, if you encounter it at mixdown, then you can often deal with it remarkably well using a specialist comb-filtering plug-in. In this example I've done exactly that, feeding the GuitarHum01_Raw file through Tone Boosters' TB_HumRemover plug-in, targeted for 60Hz and with Q setting of the comb-filter notches at an extremely narrow value of 100.
Here's a little brass riff sample, in mono. The next example files will demonstrate the two main mono-to-stereo conversion methods I mentioned in the article.
To create the side stereo effect you can hear in this audio example I multed the MonoToStereo01_MonoRaw file to two tracks, panned hard left and right. I then applied a 12ms delay on the right-panned track. The main downside of this approach is that the timing offset between the two channels causes them to phase-cancel strongly when summed, so mono-compatibility isn't terrific.
By contrast with the mono-to-stereo conversion technique demonstrated in the MonoToStereo02_StereoDelay file, the technique used to create the stereo effect in this audio example has excellent mono compatibility, albeit with a less extreme widening effect. To achieve this sound, I once again multed the MonoToStereo01_MonoRaw file to two tracks, panned hard left and right, but instead of applying a delay to one channel, I drastically EQ'd the two channels in equal and opposite directions using Slim Slow Slider's Linear Phase Graphic EQ plug-in.
This audio file contains a short sample taken from SOS July 2011's Mix Rescue remix, which I'm going to use to demonstrate some of the detailed in-sample rebalancing techniques mentioned in the article. For the sake of this demonstration, let's assume that I want to isolate the vocal in this sample as much as possible, so that I have more freedom to create my electronica arrangement around it.
In this particular track the live drum track's kick-drum was replaced with a triggered sample, so for this audio file I hunted through the remix I sampled for the VocalSample00_Raw file to find the most isolated kick-drum hit I could, and then copied onto another track under each kick-drum hit of the sample, lining up the hits as closely as I could by matching the waveforms. Flipping the polarity of the edited hits then phase-cancelled the kick-drum in the sample mix -- not enough to remove the kick-drum completely, but enough to usefully reduce its level in the mix.
Treating the file you heard in VocalSample01_KickCancel with a 24dB/octave high-pass filter at around 170Hz removed unwanted low frequencies from beneath the vocal register without altering the vocal tone -- all but removing the bass guitar and remaining kick-drum frequencies from the mix.
My next step was to use a couple of precise fader automation dips to duck the loud snare drum into the mix. However, there was only so far I could go with this before it began to pump the vocal level too overtly.
Noticing that the snare's tail was presenting a wide stereo image, I decided to try ducking it further by putting it through Voxengo's freeware MSED plug-in and then automating the Sides level downwards at the appropriate moments. Here's what it sounded like.
In order to reduce the level of the hi-hat (and further reduce the upper frequencies of the snare), I've processed the VocalSample04_SidesDucker with an instance of the Cockos ReaXcomp plug-in, limiting the frequency range above 3.5kHz at a ratio of 100:1 and attack and release times of zero and 17ms respectively. You can hear that this processing makes a big difference to the hi-hat level in particular, with only a slight loss of vocal 'air'.
As a final snare-reduction attempt, in this audio file I've applied a couple of peaking-filter cuts to the VocalSample05_SnareHatHFDucker file to target a couple of its stronger low resonances at 206Hz and 292Hz. I've kept the bandwidth of these very narrow (roughly one tenth of an octave) and also automated their gain controls to try to avoid removing too much of the body of the vocal sound into the bargain.
The final processing stage that you can hear in action in this audio file has been designed to remove as much of the rhythm-guitar strumming as possible. I've applied five super-narrow comb-filters in Voxengo's Gliss EQ (with root frequencies at 262Hz, 292Hz, 331Hz, 352Hz, and 441Hz) to individually target each of the main pitched notes in the guitar part. I automated the gain controls of each to cut with each only when required. The result is a considerable reduction in the guitar level, with only a slight alteration in the vocal tone -- I could have cut further to almost completely remove the guitar sound, but that had unacceptable side-effects in terms of rubbishing the vocal timbre, so I struck the best compromise I could. Now compare this example with the VocalSample00_Raw audio file to hear how far all the different rebalancing processes have brought us from the original sample!