Cinematic Story Explainer · compétence Open Skill

Cinematic Story Explainer

Cinematic Story Explainer est une Open Skill publique sur Digen. Ouvrez-la pour voir ce que ce workflow crée.

@skyrocketfamoffice

Manhua

8 3 9

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4023697/2026-09-05/987fnyt8.png

# Cinematic Story Explainer

## When to use
- The user wants a video and every video must follow the same cinematic story-explainer structure (dramatic pacing, story-driven, cinematic subtitle/captions, smooth transitions).
- Triggered by a topic, one-liner, idea, short script, article text, or an uploaded reference image/video/audio.
- Also use when the user wants to remake an earlier output from this skill, or advance/change only one step (e.g. rewrite the script, regenerate one beat's clip, restyle the visuals, regenerate the music, remix the final video).

## The fixed structure (applied to EVERY video, unless the user explicitly says otherwise)
Each video is a ~60 second vertical (9:16) cinematic story explainer built from exactly 6 beats, each beat = one smooth 10-second clip. The beats are always:

1. **Hook — 0–10s**: Dramatic cold open. A striking image + a bold opening line that stops the scroll.
2. **Setup / Question — 10–20s**: Orient the viewer. Who/what/when, and the central question the video will answer.
3. **Rising tension — 20–30s**: Escalation. The stakes climb; one new revealing fact or complication.
4. **Twist / Revelation — 30–40s**: The turn. The unexpected reveal that reframes everything.
5. **Payoff / Resolution — 40–50s**: The answer lands. Clear takeaway, emotionally satisfying beat.
6. **CTA / Cliffhanger — 50–60s**: A short call to action, or one lingering question that hooks the next video.

Across every beat, apply the four style pillars:
- **Dramatic pacing**: narration lines are short (~12–16 words), rhythm accelerates toward the twist, pauses before reveals.
- **Story-driven**: the topic is always told as a mini-story with a protagonist-like subject, a conflict, and a resolution — never a dry list.
- **Cinematic subtitle/captions**: on-screen caption text is short, punchy, all-caps-friendly, timed per beat; deliver as an SRT file plus a caption style note (the platform cannot burn text onto video — the SRT is applied at publishing).
- **Smooth transitions**: keep one consistent subject description and one consistent color grade across all 6 stills so the concatenated clips feel like one continuous shot; the final mix stitches the 6 clips back-to-back with no hard cuts.

## Steps

### 1. Understand the input and lock the topic
- Read the user's input (text, article, script, uploaded image/video/audio). Ask one clarifying question ONLY if truly necessary (e.g. missing topic, unclear orientation). Defaults: 9:16 portrait, ~60s total, 6 beats, voiceover narration + cinematic background music.
- If the user uploads a reference image/video, briefly inspect it (`recognize_video` for video, `get_document`-style reading where relevant) to extract the subject look and mood.

### 2. Write the cinematic script (one document)
- Build the script as a markdown document with exactly 6 beat blocks, each containing:
  - `beat` (1–6), `title` (e.g. "The Hook"), `timecode` (0–10s, 10–20s, …)
  - `narration` — one dramatic spoken line (12–16 words)
  - `visual` — a detailed cinematic image description (subject, action, light, camera)
  - `motion` — how the scene should move in the clip (slow push-in, drift, reveal…)
  - `caption` — the short on-screen subtitle text for that beat (same message as narration, condensed)
  - `subject_sheet` — one line repeated verbatim in every beat describing the main subject (face, outfit, key props) so it stays consistent
- Tool: `create_document(name="Cinematic Script — <topic>", entity_type="storyboard_scripts", content=<the script markdown>)`. Show the user the beat list and get a quick OK before generating assets (skip the OK only if the user already said "go").

### 3. Generate the 6 key visual stills (one per beat)
- For each beat, call `generate_image`:
  - `model='t2i'`, `orientation='portrait'` (9:16), `resolution='1080P'`
  - `prompt` = beat's `visual` + the `subject_sheet` + the fixed cinematic style suffix: "cinematic film still, dramatic rim lighting, anamorphic lens,