AI CREATION · AI video / Workflow / Case study
A 15-second AI film, from script to delivery
A real case study covering script, storyboard, generation, voice, edit and platform versions.
Real evidence, method and action.
This first complete edition is grounded in real production and backend records. Updated 2026-07-21. Platform metrics are dated snapshots and will change.
Backend view snapshot
15-second backend snapshot
25-second backend snapshot
Compress the story to one action
A four-year-old dreaming of the World Cup is still only a premise. Fifteen seconds can hold one clear line: chasing the ball in reality, crossing a threshold, and arriving in an imagined stadium.
Each additional shot must move either the action or the emotion. If it is only pretty, it goes.
Generation is not one click
Character, kit, age, light and direction of movement are fixed before generating each shot. The common failure is not an unattractive image; it is a person, outfit or stadium changing between seconds.
Failed clips are labelled by continuity, movement, repetition or emotional clarity so that revision has a reason instead of becoming endless random generation.
Sound completes the emotion
Narration cannot sound like documentation. Short lines, pauses, ambience and the music entry decide when the audience sees a dream and when it recognises a father preserving childhood.
Captions must work on a phone first. A child’s original voice is not required; adult narration or carefully used ambience can carry the story.
One master, several versions
The theme became both a documentary edit and an AI-film edit. In the 21 July 2026 backend snapshot, the 25-second documentary edit had about 4,081 Kuaishou views and 242 Douyin views; a 15-second AI edit had about 827 Douyin views.
This does not prove a universal winning format. It shows why the theme can be reused while the opening, duration, captions, cover and interaction are rebuilt for each platform.