How to Narrate Song Lyrics and In-World Songs in a Novel

To narrate a song or a set of lyrics inside a novel, you accept that the passage will be spoken rather than sung, then shape it so it reads as verse: keep the line breaks, add pauses at the ends of lines and between stanzas, decide whether the singing character or the narrator delivers it, and put any music around the passage instead of under the voice. AI narration has no singing voice. Once you stop fighting that, a song in a chapter becomes a pacing problem, and pacing is something you can actually control.

This comes up more than writers expect. Fantasy novels carry tavern ballads, and literary fiction quotes a hymn at a funeral. In print the reader hears whatever they want to hear. In audio, somebody has to say the words.

The constraint, stated plainly: there is no singing voice

AudioProducer.ai renders text to speech. There is no sung delivery and no melody generation, and no setting anywhere that gives a voice a key or a tempo. A character who bursts into song in your manuscript will speak those lines in the finished MP3, in the voice you assigned, at speaking pitch.

We would rather say that up front than let you find out after rendering a whole chapter. Re-generating after a change spends your word allowance again, so test a song passage on a short sample first.

What you do have is the set of controls that shape any passage: per-character voice assignment, per-line speaker tagging, per-line dialogue emotion, and pauses at several levels. That is enough to make a spoken song sound deliberate. Nothing in it will make the passage sound like music, and a passage straining toward music while clearly not singing is the worst outcome available.

Make the spoken song read as verse

A song runs on lines, and the line ending is where its rhythm lives. Default narration pauses at commas and full stops and runs straight through your line breaks, which turns four lines of verse into one long sentence that happens to rhyme.

Keep the manuscript's line breaks in the text you paste, then buy silence at the ends of them. In AudioProducer you have an inline pause you can drop anywhere, a project-wide default paragraph pause set under Edit Project, and a per-paragraph override for moments that need more breath than the default. A song usually wants the override: a longer gap between stanzas than the book uses between paragraphs, and short beats inside the stanza where a singer would breathe.

Take an invented example, the kind of thing a tavern scene might carry:

The river took my brother's boat,
the river keeps what falls.
I paid the ferryman in salt,
he never paid at all.

Read with prose timing, that lands as a run-on sentence. With a short beat after each line and a longer one after the last, it lands as a song being recited. The mechanics are covered in our guide to adding pauses and dramatic timing to AI narration, and the same instinct drives turning a poem into spoken word audio, where the whole piece is verse rather than four lines inside a chapter.

Decide who delivers the song

You have two honest options, and they sound different.

The character sings it. Tag the song lines to that character so their voice carries the passage, the same way you would tag their dialogue. This works when the song is an event in the scene: somebody performs, other people listen. Give the lines a dialogue emotion that matches the moment, since the emotion setting is per line and a lament and a drinking song are not the same read.

The narrator reports it. Leave the lines with the narrator when the song is quoted rather than performed, like an epigraph at the head of a chapter or a verse the point-of-view character remembers. The narrator's voice keeps it at a distance, which is often what the prose was doing anyway.

One thing worth knowing before you plan around it: a character entry holds one voice. If you want a character's singing to sound different from their speech, the way to get it is a second character entry with a different library voice, tagged onto the song lines by hand. Auto-Assign Characters will not do that for you, since as far as the markup is concerned both entries are the same person. For a four-line song it is usually not worth the work. For a bard whose songs run through the whole book, it can be.

Put the music around the song

The instinct is to drop a music bed under the lyrics so the passage feels sung. It rarely works. A bed under spoken verse fights the words, and the words are the only thing carrying the song.

What works better is placing the music around the passage: a bed that sits under the prose leading into the song and drops away for the lines themselves. Selection happens in the Sounds panel, and Auto-Assign Sounds will make a first pass on the scene that you then adjust. Our guides to adding background music to an audiobook and adding sound effects and music cover level-setting in more depth.

There is one more option that people miss. You can upload your own audio into your personal sound library and use it alongside the built-in tracks. If you or somebody you have worked with has actually recorded the song, that recording can sit in the render as a sound while the surrounding chapter is narrated. We do not generate it, and you are responsible for having the right to use whatever you upload, but the door is open.

Know when to cut the song down

Some songs should not survive the trip into audio at full length. Eight stanzas of invented folk ballad that a reader's eye skims in four seconds becomes ninety seconds of spoken verse, and listeners feel every one of them.

Cutting to the first stanza and a line of narration that covers the rest is a legitimate audio edit. So is replacing the quoted lyric with the scene around it: what the room does while the song runs, who stops talking to listen. The same judgment call shows up with letters, emails, and text messages in a novel, where the printed artifact and the spoken version want different lengths.

Volume is not a lever here either, so a song that gets quieter as the singer walks away has to be handled through the words and the pauses rather than through a fade. We covered that constraint in detail in our post on narrating a character who whispers or shouts.

Rights: only use words you are free to use

Everything above assumes the song is yours. If your manuscript quotes a real song, clearing that use is your responsibility as the author, and it applies to the audio edition the same as to the print one. We cannot give legal advice on it. The safe path for fiction is the one most authors already take: invent the song. An in-world ballad you wrote is yours to narrate, and it usually serves the scene better than a borrowed lyric.

The short version

Keep the line breaks, buy silence at the ends of them, pick whether the character or the narrator delivers the lines, keep music beside the passage rather than under it, and cut long songs down to what a listener will sit through. Try it on one song before you render the chapter. You get 1,200 words free with no card to find out how your verse lands.

A song is one of several page elements that is not prose. Narrating footnotes and endnotes covers the same question where the interruption is a reference rather than a verse.