video-script · Video Script Generation
Script generation across the full range from short video (7–60 seconds) to mid/long-form (1–30 minutes): shorts get three scored hook variants, second-by-second timing, subtitle copy, and cover plans; long-form gets retention optimization, rhythm breaks, forward hooks, and chapter structure.
Mode switches automatically by target length, with word counts verified by script against ~250 Chinese characters per minute. Built for Douyin, WeChat Channels, Xiaohongshu, Bilibili, YouTube, and more.
Example invocation: "Write a 30-second Douyin script on why Rust is replacing C++, educational tone."
Full brief
Positioning
video-script writes the script for one specific video: from topic, length, platform, and key points to a structured shooting script. It handles execution-level scripts — not series planning or content strategy.
Core capabilities
- Auto mode selection: switches between short-video (7–60s) and mid/long-form (1–30 min) writing by target length.
- Short-video four steps: three scored hook variants (pick three of five types: curiosity gap, pattern interrupt, identity trigger, pain resonance, authority proof) → second-by-second script → caption + cover plan → 100-point quality score (ships at ≥80).
- Long-form retention design: three-part hook, rhythm breaks every 60–90 seconds, forward hooks, chapter structure, retention-risk annotations.
- Word-count verification: all voiceover copy checked by script against length × 250 chars/min ±10% — never estimated.
- Profile awareness: with an account Profile, platform is locked, wording and tone adjusted, habitual CTA phrasing applied.
Workflow
- Provide topic and target length (required); platform, format, key points, tone (optional)
- Enter short or long mode by length
- Generate hooks → segmented script → subtitle/visual cues → caption & cover (shorts) or retention-risk map (long-form)
- Quality check: word count, hook completeness, break cadence, CTA placement
- Outputs to
outputs/: script metadata, segmented script, rhythm-break log, editing notes
Inputs & outputs
| Input | Required | Notes |
|---|---|---|
| Topic | Yes | Video topic |
| Target length | Yes | Seconds or minutes |
| Key points | Yes | Must-cover information |
| Platform | No | Inferred from length by default |
| Format | No | Educational / tutorial / review / recommendation / explainer |
| Tone | No | Professional / casual / sharp / conversational (default conversational) |
Output: full script with timecodes, voiceover, subtitles, visual cues + quality score / retention-risk map + editing notes.
Boundaries with adjacent skills
| Skill | Lane |
|---|---|
| video-script (this) | Writes the script for one specific video |
| video-strategy | Video content planning and series strategy |
| short-drama | Multi-episode drama pipeline; script-writing included |
Fit
- Second-by-second scripts for Douyin/Channels/Xiaohongshu shorts
- Retention-optimized structure for Bilibili/YouTube mid/long-form
- Content teams standardizing script quality and delivery format
Before you start
- No API key; pure text generation.
- The more specific the key points, the denser the script; platform and tone fall back to defaults when unspecified.
video-script is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.