The oldest format, the newest pipeline
Narrated stories — a voice telling a tale over images — are the most forgiving and most searchable format in AI video: no lip-sync pressure, no dialogue coverage, huge audience appetite (sleep stories, folklore, explainers, fiction channels). If you write prose at all, this is your fastest path to a finished film.
The Narrated story preset plans for it: the voiceover carries the story, every paragraph becomes a shot's narration, visuals illustrate rather than enact, shots run long (6–10s) with specified camera moves — slow push-ins, drifts, pans.
Prose in, shots out
Press Start this project below — The Cartographer's Daughter is plain prose, no screenplay formatting. Roll camera and check the breakdown: each paragraph should land in a shot's dialogue field as narration. On screen, characters don't mouth these words; the pictures breathe while the voice speaks.
Writing for narration: shorter paragraphs make better shots. If one paragraph carries two images ("the study" and "the window ritual"), split it before ingesting.
Choose the voice like you'd cast an actor
In the Export phase's Audio panel, pick the show's voice — this story wants a warm, unhurried narrator (Brian or Rachel). Then, in the Generate phase, press 🎙 Narrate all (1 credit per shot).
Listen to every line on its shot card. Reads you don't like: Re-narrate (takes vary), or adjust punctuation in the dialogue text — commas and em-dashes are how you direct a TTS performance — and narrate again. Changing the text automatically clears the stale audio.
Timing: let the pictures wait for the voice
Each shot's duration must fit its narration with air on both sides. A rough rule: narration seconds ≈ words ÷ 2.5, then add 1.5s. A 20-word paragraph wants an 8–9 second shot.
If a clip runs long after rendering, don't regenerate — trim it on the shot card (trim out-point) and the timeline recuts instantly. If it's short, either shorten the line or bump the shot's duration and re-render that one clip.
Music: the emotional floor
Generate the bed in the Audio panel: "gentle melancholic piano over soft strings, slow, wistful." Audition it live under timeline playback. Instrumental only — the negative prompt already blocks vocals, because vocals fight narration.
The multitrack render
Assemble preview cut now does real post-production in one server-side job:
- the video track — your approved takes, in order, trims applied
- the voiceover track — each narration clip placed at its shot's start
- the music track — the bed under everything from 0:00
Play the result. If a narration collides with a cut, nudge that shot's duration or trim and re-assemble — assembly is free.
Ship it everywhere
- Preview cut — your master.
- Render social cut — the 9:16 vertical for Shorts/TikTok/Reels.
- Download captions (.srt) — narration as subtitles, ready for YouTube.
- Upscale all clips (2x) before final assembly if this is going anywhere bigger than a phone.
Narrated-story checklist
- Every paragraph = one shot, one narration clip
- Shot durations fit their lines with ~1.5s of air
- Every prompt specifies a camera move (static shots read as slideshows)
- Music is instrumental and quieter than the voice
- .srt exported for the platforms that caption
You've finished the core curriculum. Go back to the Academy and pick your track's next project — or just go make something.