How to Turn a Story or Poem into an AI Video
MyDreamVid · 2026-09-12
Every writer has a drawer of pieces that deserve more than a drawer: a poem, a flash fiction, a scene from the novel that never got finished. Text-to-video makes adaptation a one-person job. The craft is not in the generating — it's in deciding what, of all your words, becomes picture.
Poems: one image per stanza
Don't film a poem line by line; film its images. Read the poem and underline the concrete pictures — in most poems there are three to six. Each becomes one shot of 5–10 seconds, and the poem itself becomes the voice-over: put the lines in quotes in the prompt of the shot they belong to, directed like a reading — a low, unhurried voice reads: "...". Between images, let the video breathe with slow cameras and quiet ambience; a poem's pacing is mostly silence.
Stories: scenes first, then shots
A story adaptation starts the way a screenwriter starts — by finding the scenes. The film workspace mirrors this: paste your story as the idea, let the AI writing steps (10 credits each, refunded on failure) propose an outline, a script, and a shot list — or skip straight to pasting your own shot list, one line per shot with a duration at the end. Either way, you edit every step; the model drafts, the author decides.
Two adaptation rules that survive contact with the model:
- Show states, not summaries. Prose says "years passed and the house emptied"; a shot list says "the same kitchen, now dusty, one chair remaining, winter light". Convert narration into visible before/after states.
- Cast it once. Generate a portrait for each character and a wide shot for each location, attach them as references, and your protagonist survives every scene change with the same face — the consistent characters guide shows how. Real people's faces are refused by every model, so characters must be invented or stylized — for fiction, that's no restriction at all.
Give the adaptation one voice
Set a style note — "ink and wash, muted colors, soft grain" — and it is appended to every shot, which is what makes twelve generated clips read as one work (styles guide). Choose the style from the text: a ghost story wants shadow and grain, a childhood memory wants warm storybook color.
What it costs
Video is billed per generated second and every shot is quoted exactly before it runs (1 credit = $0.01, credits never expire, failed shots refunded). A poem of five shots drafted at 720p on Seedance 2.0 Mini costs under $3; a short-story film of 20–30 shots is comparable to a night at the movies. Draft on Mini, regenerate the keepers on 2.5 (pricing).
Frequently Asked Questions
Should the text be read aloud in the video, or shown as captions?
Spoken almost always wins — put the lines in quotes and they are read with natural intonation you can direct. Burned-in text is the weakest skill of every video model; if you want typographic titles, add them in an editor afterwards.
Can I adapt a story I didn't write?
Only if you have the right to — your own work and public-domain texts are safe ground. The output video is yours; responsibility for the underlying text stays with you.
How long should a story video be?
Shorter than the story feels. A poem: 30–60 seconds. Flash fiction: one to three minutes (roughly 10–30 shots). Cut anything that is connective tissue in prose — video jumps between moments better than prose does.
My story has dialogue between two characters. Does that work?
Yes — quote each character's lines in the shots where they speak, and keep both characters attached as references so they stay themselves. Shot-reverse-shot (alternating close-ups) is easier for the model than two characters conversing in one frame.