Read word for word, split into beats
A pasted script is divided into beats by simple rules rather than by a model: at your line breaks first, then at sentence ends, and at commas or similar pauses if a sentence is long. Your wording is what the narrator reads.
For every beat, the AI writes a brief for the picture that sits behind it and suggests a title for the whole short. Nothing else is generated. The look you pick still shapes how the narrator delivers your lines, from a dry noir drawl to a warm gallery-guide tone.
Fit the script to the length
Narrators speak at different speeds, and each look asks for its own pace, so the number of words that fits a length changes with both. The studio keeps a live count as you type and shows the target for the length, look and voice you have chosen.
A script can run a little over the target. Past that margin it is refused before any credits are spent, with a message saying how many words it has and how many fit. Trim it, pick a longer length, or choose one of the quicker voices.
Writing for a narrator
A script that reads well on paper can sound stiff out loud. A few habits help:
- Put each idea on its own line. Line breaks become beat boundaries, so they decide where pictures change.
- Write numbers and symbols the way you want them spoken, for example “nineteen eleven” rather than “1911” if the reading matters to you.
- Leave out brackets, footnotes and stage directions; anything in the script is read aloud.
- Read it aloud once with a timer. If you run out of breath, the narrator will too.
When a topic is the better start
If you only have the idea, not the words, switch to topic mode and let the writer draft it. Plenty of people do both: generate a draft from a topic, tidy the wording, then paste the edited version back in as a script.