Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Video Narration Script (video-script)

Turns an already-analyzed video into a director-level story plan, A/V board, and timestamped narration JSON — then validates it before TTS.

Image & VideoAdvanced★ 563⑂ 105AI score 8/10Last updated: Oct 4, 2026

What it does

  • Reads a work_dir that already contains video-understanding artifacts (vlm_analysis.json, asr_result.json, agent_narration_brief.md) and walks the agent through directing → story editing → cutting → voice/narration → first-viewer check.
  • Produces recap_story_plan.json (intent, POV, change-driven beats), visual_audio_board.json (per-beat audio_owner and narration_job), optional style_card.json, clip_plan.json for cut mode, and a timestamped narration.json with emotion tags for TTS.
  • validate.py runs locally and mechanically: flags over_budget blocks, ordering errors, and rewrites measured overlaps_speech without ever rewriting your prose.
  • review.py is an advisory LLM review (hallucination/unsupported-claim detection) and is the only script that goes online, sending text excerpts to a MiMo chat endpoint — no video, frames, or audio.
  • Distinguishes CREATE / DIRECTED / REVISION modes so review notes freeze everything the user didn't mention.

Who it's for

  • Creators and studios producing recap / summary commentary videos at volume
  • Users of the zenstory video-recap automated TTS pipeline
  • Anyone trying to move narration beyond "describing the frame" toward causal, POV-driven storytelling

Examples

  1. Point it at a 90-minute drama's work_dir and ask for a 10-minute recap → it compares two viable structures, writes clip_plan.json, then output-time narration.json.
  2. After a screening: "beat 3 sounds like subtitle-speak, freeze the rest" → REVISION mode edits only style_card.json and that segment.
  3. Only need promo titles, thumbnail text, or on-screen caption fixes for a finished short → it uses the promotional-copy reference and skips the planning chain.
  4. Run python3 scripts/validate.py --work-dir <work_dir> --mode full to find over-length blocks before synthesizing voice.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/zenstory-ai/video-recap-skills/HEAD/skills/video-script/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/video-script folder from the GitHub repo zenstory-ai/video-recap-skills into my ~/.claude/skills/video-script/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/zenstory-ai/video-recap-skills.git && mkdir -p ~/.claude/skills && cp -r video-recap-skills/skills/video-script ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal and go to a working folder.
  2. Clone the repo: git clone https://github.com/zenstory-ai/video-recap-skills.git
  3. Create the skills folder: mkdir -p ~/.claude/skills
  4. Copy this skill: cp -r video-recap-skills/skills/video-script ~/.claude/skills/
  5. (Recommended) Copy the sibling analysis skills too — this skill only consumes an existing work_dir, it does not run ASR/VLM itself.
  6. For the optional review step, set export MIMO_API_KEY=your_key; to keep drafts local, skip it or set REVIEW_NARRATION=0 / pass --no-review-narration.
  7. Restart Claude Code and ask for a narration script. For confidential drafts, run only validate.py and do human review instead of review.py.
View source on GitHub ↗License: MIT