This is a fairly impractical post.  How many people want to write the words to an opera?  Perhaps the first rule of writing libretto would be, simply,

1) Don’t. 

Nonetheless, I’ve been asked a lot recently how one goes about doing it — and so below are some guiding principles, learned from experience:

2) Stick to an 8th-grade vocabulary.  Singing shines a spotlight on words, and some words crumple under the heat.  As a rule, the rarer the word, the easier it crumples.  You’ve probably heard the old adage, “Never use a $10 word when a $1 word will do.”  When your text is put under the spotlight of singing, the embarrassing effect of a $10 word is multiplied exponentially.  

(To this day I’m not sure how this singing “spotlight” works — I only know that there are times when the choice of a word seems perfectly right sitting at my desk and reading it aloud to myself — but later, when a trained singer delivers the line in person, it makes that same word seem, for lack of a better word, goofy.  In these cases I always replace it with a simpler word.  Perhaps this phenomenon has to do with how much of the ear is occupied listening to music.  When we’re listening to music, there’s no extra bandwidth in the ear to go off and riffle through the occipital dictionary and get comfortable with a high-octane word.  When music is sounding, fancy words — even ones we know — seem flown in from farther away.)

Likewise Mark Twain’s famous advice, “If you find an adjective, kill it,” is doubly important for sung text.  This is not a matter of taste, but of simple mechanics: in sung text, the music is already supplying the adjective.  A character should almost never tell us that they’re feeling hopeful, resolute, wistful, apoplectic, etc..  The music should show those emotions directly while the character says whatever they need to say.

3) Have characters talk about offstage stuff at your own peril.  Text about offstage action is a special effect, and should be used judiciously.  

I imagine a spectrum, where verb-driven statements addressed from character-to-character (“Get out of here!” “Hand it over!” “Take me now!” ) are at one end, and description-driven statements addressed from character-to-audience (“I remember when I was a kid, I used to play with this green glass ashtray…”) are at the other:


The best librettos sit toward the left side of the spectrum.  

Obviously, there are no hard-and-fast rules here.  Having a character speak about something the audience hasn’t seen onstage can be enormously powerful, especially if you’ve starved your audience of such “offstage data” verbal tissue for a while (counterintuitively, the more specific the detail in such moments, the more poignant the effect).  

To myself, I call this poignant effect the “Grey’s Anatomy Principle.”  In nearly every episode of Grey’s Anatomy, at an emotionally climactic moment, the screenwriters do this weirdly-specific-offstage-detail trick to shoehorn our attention inside a character’s experience:

Character (out of nowhere): The green glass ashtray…

Other Character (silently, to themselves): WTF?

Character: The night before he died, I broke my father’s green glass ashtray…

And then some beautiful, crucial character stuff that’s been hidden in the slingshot for 45 minutes gets revealed.  The weirdly personal, offstage detail can actually amplify the big, universal onstage truth — but you need to have set up all your big universal onstage truths (i.e., your character motivations) with crystal clarity beforehand in order for this to work.

I would describe the use of offstage-action text in the same way I once heard the composer Joan Tower describe the use of crash cymbals: “It’s like eyeliner.  A little goes a long way.”

4) Short grammar beats long grammar.  Singing necessarily involves slowing down the rate of verbal delivery.  Given any possible English sentence, it takes longer to sing it in a natural way than it does to say it in a natural way.  This means that a short grammatical structure will carry more power than a long grammatical structure.  Faulkner’s sentences are way cooler than Hemingway’s — but when writing libretto, go for Hemingway.

I’m often amazed, whenever I’m moved by a sung line, how grammatically impoverished it is when stripped of the music.  My favorite example is the verse line in “Landslide” by Fleetwood Mac:

But time makes you bolder.

Even children get older.

And I’m getting older, too.

When you write it out… there’s really nothing there.  Can you imagine how apocalyptically high you would have to be to imagine that these lines constituted a great poem?  It’s just stray, grammatically unconnected observations — and painfully obvious ones at that.  Yet add in Stevie Nicks’ perfectly tarnished voice, and the uncanny symmetry of that first-inversion chord floating in the guitar, and the way those lines gently distend the verse structure, and it’s unspeakably profound.  Those lines waste me every time —and yet to look at the text alone, I can’t figure out why.

Another great example is the beginning of “The A-Team” by Ed Sheeran:

White lips,

Pale face

Breathing in


Burnt lungs, 

Sour taste.

See how there’s no grammar connecting those images?  They’re just loose objects, and only the music holds them together.  Yet they’re far more evocative like this than they would be if the text were trying to micromanage our journey through the scene with unnecessary connecting grammar.  

Another way to think of it this is: the music is the grammar, the forward-moving connecting principle.  If the text insists too loudly its own linear grammatical trip, we’ll end up hearing a faint sense of disagreement between the words’ version of that trip and the music’s.

5) Mind the Sondheim Blinders.  I’m sure he said it more clearly, but I once heard an interview with Stephen Sondheim in which he noted that the hardest thing about writing sung text is that it can never be chewed on and reconsidered as can text in other literary genres.  An obscure line at the beginning of a poem, or a detail passed over in a novel, can be flipped-back-to and reviewed.  Indeed, the flipping-back-to-and-reviewing, the reading and ever-deepening re-reading, is one of the primary joys of these genres.  Such text can be treated like a three-dimensional object: saved, examined, turned over, reconsidered after a good night’s sleep, lived with for a while.  

Not so with sung text.  Listening to a libretto is like reading a page where an eraser is always following just behind your eye, obliterating everything that’s come before.  You’re left only with what you can remember — and you can’t even stop to think about that, because you’re still busy hearing new lines and new music.

This means that grammatical antecedents (see #4), and informational details (see #3) — and anything at all which has been sung more than 10 seconds ago — might as well not exist.  Everything you write has to be clear the first and only time it’s heard. Nothing you write can depend on anything besides itself.  You have to imagine hearing every line afresh, in an amnesiac vacuum.

To mistranslate, then agree with Derrida: There is nothing outside the text.  

Not even, in this case, the text.

6) Skin is soft, bones are hard.  Sometimes when I listen to people discuss the interaction between text and music, I notice they’re talking exclusively about how text and music interact on a local, beat-by-beat, rhythmic level — “this fits, this doesn’t.”  They are concerned only with the surface-level interaction of text & music, or what I would call the “skin” of the operatic beast.   This includes prosody, but also extends to questions of vowel path and vocal range and color.  Of course, it’s nice if words and music match at this “skin” level, but it’s not the most important thing.

Deeper, at the bone, there is a level at which the music and the text must match absolutely: character intent (text) must match large harmonic blocking (music).   This is the crucial contact point between the two.  If your character is saying something for which the motivation is X, the music — specifically, the harmonic map — had better be mirroring that character’s interior X or else productively countering it.  This contact point between text and music is the one audiences actually hear.  If this bone-level contact is occurring, you can cut a word, add a measure, change figures in the vocal part, do whatever you will on the skin-level — and the product will still work.

Which is good news.  When we secure the interaction of words and music at the bone level, it turns out we can adjust things at the skin level however we like.  The skin of the opera is malleable, simply because word choice and rhythmic declamation in the voice part should be malleable until very late in the game.  Both composer and librettist should be flexible and creative enough spin out multiple solutions at the skin level, always with the aim of showing the bone structure more clearly.

7) Your libretto shouldn’t make complete sense.  One way I know I’m off base is if I look at a drafted passage and I can see everything in the text.  Meaning: opera is most successful when the text leaves the music something to do, some part of the narrative weight to pull.  If the text conveys everything by itself, it will jam up the flow of the music like logs in a stream.  If the text says it all, why have an expensive union orchestra on hand?  Why have costumes?  Why props?  Why lighting design?  Why, for that matter, professional actors, who can convey a thousand words in a glance?

Someone once asked my wife Jennifer, a collaborative pianist: “Is great art song more about the text, or more about the music?”  Her reply was characteristic in its improbably true precision: 

“Great art song is 51% music.”

And I’d say exactly the same thing about opera.  The text has to know where those crucial moments are when the music alone gets to tell the story.  This means that, when you look at a good libretto on the page, it should have some holes in it.  That extra 1% — that moment when the text allows the music to cross the finish line by itself — is the magic of opera, and as a librettist you have to make room for it.

Writing libretto is like setting dominoes up on their ends, row after curling row, ready to be knocked over.  Writing music is like knocking them over.  Unless the dominoes are perfectly set up, you can’t knock them over and get that thrilling effect we all love.  But it’s the happening itself — the falling — that we love to watch, not the perfectly set-up dominoes standing still This is the 1% that libretto-writing has to make room for: the musical tip of the finger, the magical sense of forward action, the illusion that the dominoes are doing something all by themselves.

Often I hear about new opera projects in which part of the development process is a public reading of the finished libretto… before any music is composed.  Such a process is — to politely understate the matter — criminally insane.  Perfecting the libretto to the point that it could be publicly performed by itself will almost certainly wick material over into the libretto that should be left to the music.  

Moreover, the librettist would have to be psychic to know exactly what the musical form will need from him or her months before the musical form exists.  The librettist will unwittingly be making deep-tissue structural decisions for the composer.  Of course, there are times when the librettist should do exactly that.  But the stream of influence between text and music should at least be able to flow both ways.

Some people say opera is a dramatic form with musical accompaniment.  I’m not sure I’d go so far as to say it’s a musical form with dramatic accompaniment, but I’d definitely say: it’s 51% music.

Which leads us to the #1 all-time most reliable rule for writing a great libretto:

8) Work with a great composer.