---
name: video-script-templates
description: Reusable video script templates for agents: two-column AV scripts with the A-roll and B-roll outline method, word-count to runtime math, and teleprompter-ready formatting. Use when writing, timing, or formatting a script for YouTube, short-form, explainer, or training videos.
---

# Video Script Templates

## Purpose

Turn a topic into a shoot-ready video script without inventing the structure from scratch. Provides four proven templates, the A-roll/B-roll outline method, runtime math, and teleprompter formatting, plus a CLI that times any draft.

## Workflow

1. **Pick the template.** Short-form (Hook + points + CTA), long-form/explainer (two-column AV), talking head (single-column prompter copy), training (sectioned AV). See `references/templates.md`.

2. **Write the audio spine first (the A-roll/B-roll outline method).** Draft the spoken words as one continuous conversational pass, before planning any visuals. This is the A-roll spine: what the host says, or the voiceover, or the interview answers.

3. **Plan the visuals beat by beat.** For each beat of the spine, fill the VIDEO column: keep the speaker on camera (A-roll) or cut away (B-roll). Rule of thumb from working scripts: a visual change every 2-3 seconds for short-form, every 5-8 seconds for long-form. One idea per shot; new idea means new visual.

4. **Time it with real math.** Run `python3 bin/script_timer.py draft.txt --wpm 150 --target-seconds 120`. Adjust word count until it fits. See `references/pacing-table.md` for pace presets and quick word budgets.

5. **Format for the teleprompter.** Convert the final spoken copy with `python3 bin/prompter_format.py draft.txt > prompter.txt`. See `references/prompter-format.md` for the rules it applies.

6. **Read it aloud.** Every stumble is a rewrite order. If the mouth trips, the ear already fell.

## Output Contract

A finished script package contains, at minimum:

- `script-av.md`: the two-column script (timecode, video, audio, on-screen text) or the chosen template filled in
- `prompter.txt`: teleprompter-ready single-column copy (for any on-camera delivery)
- A stated target runtime, a word count, and the WPM assumption used to check it
- A B-roll/shoot list grouped by setup so shoot day batches efficiently

## Operating Rules

1. Never invent numbers, dates, quotes, or events. Unverifiable facts get cut or framed as opinion.
2. Hook lands in the first 3 seconds for short-form, first 5-10 for long-form. No throat-clearing openers ("in this video I'm going to", "welcome back to the channel").
3. Write copy the way a person talks: contractions, short sentences, one idea per line of prompter copy.
4. On-screen text reinforces key numbers and phrases only, never full dialogue.
5. Every row of the AV script makes the audio state explicit. If the editor has to guess what the audience hears, the row failed.
6. The cheapest edit is on paper. Cut words to fit the runtime before the shoot, not in post.
7. Keep SKILL.md-style filler out of client scripts: no "delve", no "tapestry", no "it's not X, it's Y" pivots, no hollow inspirational filler.