# scripting-and-storyboarding The **pre-production system** — start from the spine, write both columns, envision every beat, number the shots, and edit on paper first. The agent writes the plan; the **human shoots and judges**; **WoopSocial publishes** the finished video. (Craft skill — no tool file.) ## The POV: videos are won in pre-production — the cheapest edit is on paper Chaotic shoots and endless edits are a pre-production deficit, not a talent deficit: the plan is "where you catch problems before they become expensive fixes on set or in post." Four truths carry this skill. **(1) A words-only script plans half the video** — the visual half gets improvised, and improvisation defaults to a static talking face; the **two-column AV script** (audio | visual) writes what's *seen*, beat by beat, so the change-every-few-seconds rhythm is planned, not hoped for. **(2) The storyboard is a decision document, not art** — its only question is "what's on screen at this beat?", answered by a shot table, stick figures, or AI previz frames (2026 pipelines make character-consistent boards in minutes — the tested benchmark: 12 frames from ~3 hours to 40–60 minutes, with explicit camera intent and a character bible as the craft rules). **(3) The shot list regrouped BY SETUP is the batching trick** — two weeks of content in one afternoon comes from sorting shots by location/framing/outfit, never by video. **(4) Runtime math doesn't negotiate** — ~130–150 wpm means a 300-word script is a 2-minute video; the stopwatch pass and the paper cut happen *before* the shoot. And the 2026 twist: **AI video sequences need more pre-production, not less** — every panel becomes a generation's shot brief, and previz frames cost cents where generations cost credits. ## Read these first 1. **idea-generation-and-ideation** — the idea being productionized. 2. **short-form-video-script** (or the long-form skill) — the retention spine this turns into production. 3. **brand-profile** + **design-and-templates** — visual language, end cards. ## The framework: SCENE (Depth: `references/the-scene-framework.md`.) - **S — Start from the spine:** one job, audience, format target, beat outline — before any words. - **C — Columns: audio + visual:** the two-column AV script with [VO]/[SFX]/[Action] tags; if the visual isn't written, it will be improvised; no "me talking" twice in a row. - **E — Envision every beat:** the cheapest board that answers "what's on screen" — shot table → stick figures → AI previz frames (explicit camera intent, character bible + ref image, micro-beats, directional not final); skip boards honestly where a shot list suffices; for AI video, each panel = the shot brief. - **N — Number the shots:** shot list with setups → **regroup by setup** for batching (sequence setups, buffer takes on hooks, label footage `vid#-shot#`); runtime math + feasibility pass. - **E — Edit on paper first:** table read with a stopwatch, cut the middle not the spine, the honesty pass (no staged-as-candid, no lifted scripts) — then hand off to the shoot, the edit, and WoopSocial. ## The reality (verify-quarterly) Stable craft: the two-column AV script (broadcast standard), ~130–150 wpm runtime math, the storyboard-as-decision-document, the by-setup batch regroup. The 2026 AI layer (attribute): script→board pipelines matured (Boords — free tier, sign-off layer; Storyboarder.ai — 250K+ creators, animatics; LTX, mStudio, Studiovity; free Wonderunit); character consistency is now baseline; the practitioner benchmark cut 12-frame boards from ~3 hrs to 40–60 min using LLM-beats → visual directives → seeded frames — with tags ([VO]/[SFX]/[Action]), explicit camera intent (*"the AI will guess — don't make it"*), locked palette, and a character bible; boards stay directional; traditional illustration runs ~$50–300/frame (the economics behind the shift). AI-video sequences: panel = shot brief (luma's DREAM); board first, generate second. **Attribute all; verify-quarterly.** Full detail: `references/scripting-and-storyboarding-2026-reality.md`; the templates and two worked examples: `references/templates-and-examples.md`. ## Honest scope (never violate) - **The agent** writes every planning artifact (beats, AV script, board/briefs, shot list, batch plan, timing pass, footage map); the **human** shoots/generates, judges every take, and approves — the agent cannot see footage or operate a camera and never fabricates "that take works." **AI previz** follows the image rules (no unpermitted likeness; original/consented characters; disclosure if frames publish). - **Production integrity:** no staged-as-candid content (actors posing as unaffiliated strangers = deceptive endorsement; skits are fine **disclosed**); no lifted scripts (structure study yes, verbatim no); honest runtime/feasibility. **WoopSocial publishes** the finished video; it does not script, storyboard, shoot, or edit. (Full scope: `references/scope-and-connections.md`.) ## Distinct from its siblings (route correctly) **scripting-and-storyboarding (this)** = the pre-production system · **short-form-video-script** = the retention-words craft (WATCH writes the spine; SCENE productionizes it) · **talking-head-and-piece-to-camera** = the on-camera delivery · **capcut / descript** = the edit executing the AV script's visual plan · **luma / ai-video** = the generations whose shot briefs the panels become · **flux / image-prompt** = previz frames + reference stills · **storytelling-and-narrative** = the narrative angle/WHAT this schedules into shots · **idea-generation-and-ideation** = supplies the idea. ## Where this connects Reads first: **idea-generation-and-ideation** + **short-form-video-script**/long-form + **brand-profile** + **design-and-templates.** Feeds: the shoot day (the human), **talking-head-and-piece-to-camera**, **luma** (shot briefs), **capcut/descript** (footage map + AV script), **content-calendar** (the batch). Publishes via: the finished video → **scheduling-and-queue → WoopSocial.** Measure with: shoot efficiency + edit speed + published retention via **analytics-and-reporting** — never fabricated. ## Definition of done A shootable, editable plan: the spine confirmed (one job, beat outline from the script skill), the two-column AV script written with every visual beat specified ([VO]/[SFX]/[Action] tags; no "me talking" twice in a row), the storyboard produced at the cheapest sufficient fidelity (shot table, stick figures, or AI previz frames with explicit camera intent + a character bible — directional, never final art; skipped honestly where a shot list suffices), a numbered shot list regrouped by setup with the batch plan, buffer takes, and a labeled footage map, the runtime verified by stopwatch math (~130–150 wpm) and cut on paper before the shoot, and the honesty pass held (no staged-as-candid, no lifted scripts, feasibility stated straight); AI-video panels written as generation shot briefs with previz-before-credits economics; the human shooting and judging, the edit receiving a clean handoff, and the finished video publishing via WoopSocial; **nothing staged as real, nothing plagiarized, no fabricated production claims**; and correctly distinguished from short-form-video-script, talking-head-and-piece-to-camera, capcut/descript, and luma. --- 资料：Frank Heijdenrijk / pavelhov/social-media-skills 原始版本：https://github.com/pavelhov/social-media-skills/blob/1680b37bf0dc32b52da247a8196ef27a6cd47bd4/skills/scripting-and-storyboarding/SKILL.md 许可：MIT。目录与内部链接已经整理，未收录图片已省略。 MIT License Copyright (c) 2026 Frank Heijdenrijk Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.