
Before I go any further, I do want to point out that most pop/rock/R&B songs don't need this treatment, so just keep on doing that Maroon 5 song at 105 BPM from beginning to end. It's OK. For the times you need to ebb and flow, though, you'll need to massage what's called the "tempo map" of your MIDI file.
Most MIDI sequencers/notation programs/Digital Audio Workstations have functions where you can stretch and squash tempo and meter like Silly Putty. You just may not have discovered it yet, but chances are you'll find them if you poke around a bit. Another good thing about MIDI is that it is much more universal than the black art of recording a cappella, so it's very likely that you can find someone to help you put a track or two together so you can learn the tricks faster. Check your local college's music program, especially people studying film scoring and the like - they are using these functions all the time, so bribing a student with a free lunch, or even a palm full of small bills is well worth the outlay as it will make your recording life FAR easier in your upcoming project.
{mosimage}
Example 1 shows a typical tempo operation where a ritard is happening over 2 measures (this happens to be the sequencer in Pro Tools, but most others work similarly.) You can usually control the shape and degree of the slope as well to get the feel just right. Experiment a bit with these functions and soon you will be a MIDI Master!
I'd like to take a step back now and go back to the beginning - to the beginning of the song, that is. How do you usually start a song? Well, most of the time you blow a pitch, count in the tempo, and everyone starts singing. All well and good, but usually it takes a few notes or measures for everyone to find their voices, let alone the general pitch and groove! You know how sometimes if you haven't sung a note for a minute or two, it seems like your voice forgot how to land pitches and your attacks sound like a bagpipe? Recording gives you a second chance that you would never get live, so I'd like to divert for a second and talk about the "false start":
Here is an audio example of the beginning of a cue track for "Star Spangled Banner". We hear a chord to start then get a practice run of the first line. This is where people can find their voices, get into the groove, and take a nice rhythmic breath into the start of the song that will actually be on the final recording. Listen to the example, sing along, but cut the end of (the lyrics) “dawn's early light" short. Staying in tempo, take a breath and start that line again with the cue track. You'll find that the second time though you are most likely more accurate in pitch and rhythm, as well as more relaxed and able to concentrate more on performance rather than trying to find the tempo again.
You may notice that the rhythm (hi-hat) cue is in sixteenth notes rather than just counting the quarter notes of a standard "click". I've found that this is easier to follow, as there is really a lot of time happening between quarter note clicks and people tend to start guessing where the next attack or release is going to happen. You can program any sort of sub-beat rhythm you want, whether it be full drum kit or just ticks, but being able to actually have it drive the singers is key. One technique I came up with years ago was to have each sub-beat on its own channel (if you can spare 4 channels for your click track!). This way you can change the feel of the click on the fly using fader moves depending on the part being recorded, etc. Hard to explain in words, but once you try it, there's no going back! I'll try to explain more below:
{mosimage}
In example 2 (you can click on the image to see a bigger version) you can see many things going on here in the ProTools session. The top ruler represents bar numbers, and below that the narrow pink bar holds all of the tempo information. You can see the tempo dropping from 85 to 70 from bar 27 to 29, which is represented by the tempo operation from example 1 mentioned earlier. You can also hear an audio example of what you see in ex. 2. Sometimes, as in the case of a fermata ("banner yet wave" at measure 29) just stretching the time isn't enough. You'll see in the green bar underneath the tempo that we added a couple extra beats by changing the 3/4 bar to 5/4. This is stuff that you're going to need to do by feel, just by messing with things until they match your interpretation of where the time should be. Below the lyric markers, you can see an example of my 4-channel click, which are basically high hats offset by sixteenth notes. Different combinations of levels can make big differences in feel. Generally, as in this example, I have the downbeat up loudest, followed by the backbeat eighth notes (3rd track) and then the sixteenths around that down even lower. If this were a more straight ahead rock song, I might turn up the 3rd (eighth note) track a bit more for the basses to give them more pulse, while using the accented sixteenth groove to give the guitars more swing.
Here is another audio example of a straightforward rock click track ("Roll To Me" from the Stanford Mendicants' upcoming album) where I am randomly moving the 4 faders to demonstrate the changes in feel. Of course I wouldn't do it like this while they were singing, but you might be able to tell how it should benefit different voice sections at different parts of the song. Sing along!
Wow, now I just seem to be talking nonsense... did I have a point?
Just take a listen to the audio example, which on its own it might sound a little strange, but you can see how you would be able to find each attack and cutoff and place each consonant with confidence, all the while sounding like you weren't singing to a click track!