Turn a finished script into a narrated AI video

01 / Full Video workflow

Start with narration, not a collection of prompts

Paste only the words you intend the audience to hear into Finished script. Keep research notes, pronunciation guidance and visual ideas separate so they do not accidentally become spoken narration. Choose the narration language and landscape, portrait or square format before planning.

Scripts within the supported length keep their wording. Longer scripts may be adapted to the 20-minute target: compare the proposed narration with your original before you approve media generation. Actual voice timing can differ from an estimated reading time.

• Read the script aloud once and remove sentences that depend on unseen context.
• Spell out ambiguous numbers and acronyms when pronunciation matters.
• Verify facts and obtain permission for quoted material and references.

02 / Full Video workflow

Make one visual decision per narrative beat

A useful scene makes the current sentence easier to follow. Change the image when the subject or point changes, not simply because a few seconds have passed. Use consistent character descriptions and visual style where the same subject recurs.

The following scene map is an illustrative planning example, not a claim that these three shots have been generated.

03 / Full Video workflow

Review characters and scenes before committing to the whole export

Check recurring character references first, then the storyboard. Inspect anatomy, object continuity, unintended text and visual claims. Regenerate only the asset that needs work when possible; reusing an acceptable asset avoids repeating its generation.

Optionally animate selected scenes or all eligible scenes with Fast (480p) or Quality (720p) five-second silent clips, in 16:9 or 9:16. Motion is slowed to fit longer narration; this is not continuous character animation or lip-sync. Comic and square scenes are not supported by scene animation. Both public animation options support eligible Relax Mode with the same configured wait and lower priority as other Relax operations, without spending account credits.

Choose visual direction before generating references: cartoon for a narrated fable, anime for an illustrated character story, or horror for non-graphic suspense. These presets guide appearance; they do not guarantee continuity, frame-by-frame animation or rights to existing characters.

04 / Full Video workflow

Listen to the voice and check the subtitles together

Choose a narration voice and listen for pronunciation, names, pacing and sentence endings. Match the displayed script to what is actually spoken. Subtitle timing follows the measured audio; a requested duration is a planning target rather than a promise of an exact final runtime.

Preview at phone size. Keep subtitles legible against light and dark parts of the image. Add background music only if it helps the narration, and check that the mix does not obscure words.

05 / Full Video workflow

Inspect a real export and its limits

The Island That Kept the Rain runs for 10 min 10 sec, with 42 reviewed scenes, 10 short AI-animated shots, English narration, music and subtitles. Its storyboard was manually corrected after an unintended human reference appeared in an earlier attempt. The 1231 credits include that discarded attempt and revisions. Inspect its public run record; this is a curated credit-funded example, not continuous animation or a Relax latency test.

The Rooftop Garden is a fresh 66.54-second production example made from an original finished script. It contains eight new illustrated scenes, English narration, music and subtitles. All final images were reviewed and three were regenerated. The account consumed 95 credits including revisions; the run record preserves prompts, review decisions and original-file hashes. This curated credit-funded result is not a Relax or 20-minute test.

Our published 11.87-second example runs the production Full Video renderer on existing public editorial inputs. It includes two narration scenes, camera movement, subtitles, mixed music and the ending. The run record identifies inputs, output hashes and measured render time.

That earlier render-only sample does not demonstrate new generation, a Relax wait or an end-to-end 20-minute project. Its zero new credit charge reflects reuse, not a claim that every faster-generation workflow costs zero.

06 / Full Video workflow

Understand free access, export and rights

Free welcome or earned credits; automatic lower-priority Relax when credits are exhausted and no surveys are available, subject to eligibility, fair use and free capacity. Optional paid credits provide faster processing. No subscription is required.

Each operation joins the normal lower-priority Relax queue. Media within the same admitted operation share its initial wait, then process sequentially with fresh capacity and daily-budget checks. Separate operations and regenerations have their own admission and wait. Completion time varies; long projects can take days. Export the finished 720p MP4 or an asset ZIP. Private exports remain downloadable for 72 hours; save a local copy.

Commercial use is permitted under Donatello's Terms, subject to input and third-party rights. Copyright, exclusivity and platform monetization approval are not guaranteed.