Sung lines and read copy are not dialogue
Dialogue is one register among several. A sung line has fixed timing imposed from outside, and read copy has a rhythm that belongs to the medium rather than to a character. Both behave differently from a conversation. As of 2026-09-12.
| Register | What sets the timing | What breaks first |
|---|---|---|
| Dialogue | The shot length, and the writing | Pacing, on long lines |
| Read copy | A slot, usually to the second | Emphasis, which lands generically |
| A sung line | A bar, imposed from outside | Pitch and sustain, not the words |
Inclusion rule. Registers of spoken delivery a production might ask for, compared by what constrains each. Order. Fixed order: dialogue first, then the two registers it gets confused with.
1External timing is a different constraint from a shot ceiling
A line of dialogue has to fit a shot; a sung line has to fit a bar, and the bar was decided by music that already exists. That inverts the usual order, because the audio is the fixed element and everything else is fitted to it.
One entry in this register documents a route for exactly that: audio-to-video, where an existing track drives the generation. It is the only published arrangement here that suits a musical number.
2Read copy fails in a way dialogue does not
Advertising copy is written to be emphasised, and emphasis is the property a synthesised read handles worst. A line can be perfectly intelligible, perfectly timed and completely flat, which is fine for an explainer and fatal for a sell line.
Nothing in this register offers a lever for it. Rewriting the copy so the emphasis falls on sentence structure rather than on delivery is the only approach available.
3Nobody here names a register
Every reading in this register describes speech. None of them names singing or read copy as a kind of audio the product expects, so whether a musical number or a sell line is in scope is a question for a test and not for a page.
That is worth knowing before a season is planned around either. A producer asking whether a series can ship in German gets an answer from some entries; a producer asking whether a sung scene is possible gets one only from the audio-driven route.
4Where the published detail is logged
What is described above is how the techniques work in general. Which models document which of them, in whose words, is kept in the speech controls table with a date on every field.
- A second control — four levers documented once each
- The language column — where each entry's language answer sits
A mechanism note rather than a documented field. Nothing here is attributed to a model, and nothing here is a claim about one. The sourced material is on the speech controls table. Related: First tests, Subtitles and length.
5How one tool handles speech
For the speech side of “Singing and read copy”, SceneMixer's guides describe how lines are spoken:
The new name is written everywhere the old one stood: storyboard lines, episode titles and summaries, the scene lists and already-parsed dialogue.
SceneMixer, Reusing a Character, Scene or Voice Across Projects — “What happens to the old name in the storyboard?” (read 2026-10-13)
The project has one dialogue language; lines you write in another language are spoken as written after a confirmation.
SceneMixer, Dialogue language and video tier — “Can characters speak different languages in one episode?” (read 2026-10-13)