Adding Feel with Digital Audio
Although some people seem to forget this, the easiest way to record a human feel is to hire a good musician. If that's not an option, there are many ways you can mess around with timing.
First, when recording drums, allocate a separate track to each drum sound if possible. (If you don't have the track count, consider a mix of MIDI and digital audio, such as using digital audio for cymbals and other sounds that don't sound realistic if the same sample is being triggered over and over, and MIDI for kick and toms.) In a digital audio editor, this allows you to slide individual drum parts back and forth in time simply by grabbing a track and moving it. Even slight changes can make a big difference -- 5 milliseconds (ms) is a significant shift in feel-land, and it's unlikely that you'll exceed ±10ms.
Sliding a track is a fairly global maneuver; many times, you'll just want to move one or two strategic hits. For example, in the tune "Paradise" from my last CD, there's a section with a fairly steady orchestral drum pattern. Then voices come in and sustain while the drum drops out. When the drum re-enters four beats later, the downbeat hits just a tiny bit late. As a result, the brain expects the beat to fall in a particular place, but is fooled by having it hit a little later than expected. This adds interest, and also builds up the tension a bit just prior to the drum's re-entry. (You can hear this example, as well as some others showing the effects of "feel" edits, at the top of this page.)
Many programs (Digidesign Pro Tools, IQS Samplitude, Emagic Logic Audio, Steinberg Cubase Audio, etc.) allow user-placeable/editable snap points. Editable snap points make it easy to change the feel by placing the point a little later or earlier than the beat. The advantage of snapping to a point rather than just sliding audio around is that it simplifies replacing the sound later on, as well as providing a template if you want several parts to hit at precisely the same time.Tempo Time
One difficulty with digital audio arises when you need to change the tempo after something has been recorded, the way you can easily do with a MIDI sequencer. You have two options: variable speed, which lengthens or shortens the musical material to change tempo but also changes pitch, and time compression/ expansion ("time stretch") algorithms, which change length but don't change pitch. Unfortunately, these algorithms also affect sound quality, and tend to give more objectionable results when processing stereo program material.
The best solution is to vary both pitch and duration before recording your melodic parts, as "varispeed" algorithms are kinder than time-stretch algorithms to drums and percussion. Algorithms that change duration without changing pitch are another matter. Of course, the less you stretch, the better. For the slight tempo tweaks that are usually sufficient to add feel -- say, upping the tempo during a solo by 1 bpm, and dropping it down 1 bpm during the chorus -- most algorithms will be satisfactory.
As the amount of stretching increases, quality decreases. Some algorithms (such as the ones in Sound Forge) give you several options so you can trade off timing accuracy, minimum "warble," best timbral preservation, etc. Others are more "take it or leave it," and may be optimized for percussion or melodic signals, but not both.
Because I capture parts to digital tape, it's easy for me to bounce the drum tracks over to a hard disk recorder (via the analog I/O) while varying the tape speed, then bounce back to tape via digital I/O as final storage for the audio. With hard disk recording by itself, you typically define a section of audio and apply some DSP (onboard or third-party plug-ins) to change pitch or do time stretches.