1. Your musical phrases and accents should "tile" the dialogue like a brick wall, not be coincidental.
As in the picture above, notice how the beginning and ending of each brick occurs in the middle of the surrounding bricks - not at the same time. This creates the most cohesive structure. Music and dialogue react the same way.
If one imagines that the dialogue is the equivalent of a lead vocal on a pop arrangement, this makes perfect sense. The musical accents go in the spaces and answer the lead vocal. The flourishes leave space for the words to be heard.
As an example, I always reference this Nelson Riddle arrangement of the Cole Porter tune, "Night and Day."
2. While the dialogue is being spoken, do not make any drastic changes to the music. Make your changes in the spaces.
To understand this concept, you have to first imagine how the brain listens to dialogue and music together.
Consider: The brain has a wonderful way of compartmentalizing what it hears so you can sit in a noisy environment and understand the person standing in front of you, even though there may be countless other distractions. As long as the background noise is somewhat homogenous, the conversation goes uninterrupted. It's only the occasional boom or jackhammer that can disrupt the flow.
It's the same with music under dialogue. Once the brain understands what the music is doing, and how it is functioning in the acoustical environment, you have no trouble understanding the actors in the middle of the action. But the second that boom or jackhammer starts (or the musical equivalent), all attention is drawn away from the dialogue. So what would these boom or jackhammers be? Here's a partial list of things to avoid:
- Percussive accents (this seems obvious).
- Drastic changes in instrumentation.
- Changes of tessitura, i.e. the range the instruments are playing in, relative to their overall range.
- Changes of tonal root or key center.
- Drastic changes of tempo.
- Anything that obviously calls attention to itself.
3. Melodies should not be avoided, but should be treated with extreme care.
Going back to the concept of the brain and how it compartmentalizes sound, one can imagine how a melody will best interact with dialogue. As long as the brain understands the melody, and that it doesn't draw undue attention to itself while the dialogue is spoken, it can enhance the audience experience. That said, here are a few things to avoid in your melodies (you should begin to see a pattern here).
- Melodies should not begin or end while the dialogue is being spoken.
- If the melody has any large skips in range, try to place those in the silences.
- If the melody has any distinctive appoggiatures or accents, look to put those in the spaces.
- Basically, anything that causes the focus to shift to the melody, while the dialog is being spoken, should be avoided.
4. Always monitor the dialogue while you are writing the music and treat it as a dub mixer would treat it!
My first step when scoring a dialogue scene is importing and treating the dialogue as a dub mixer would treat the dialogue - with appropriate noise reduction, EQ, compression and a serious mix pass. This is done before the first note of music is written.
Then, as I compose, I am always checking to make certain every line is completely understood - that nothing I am creating is conflicting. This is truly the best way to know for certain that you are not stepping on the words.
There is one other added benefit to this process of treating the dialogue in this manner:
When the director/producer finally does hear the music you have written, and the dialogue is noise free and cuts through the background, it is often the first time the director has ever heard the words properly presented - as they will be in the final mix. Because of this, I can't count the number of times I have heard my clients say, not knowing how much effort I put into the dialogue, "Man, the words always sound so much better with your music behind it then when I just hear it in the AVID."
If they only knew... Let's let this be our little secret, shall we?