SkillFoxx

h3-storyboard skill

A Claude Code skill that breaks a script into shot lists for MiniMax H3 and teaches facial direction that renders.

  • 196 ★ on the repository
  • Last commit
  • MIT
Skill

Minimal access

Why this level

  • The skill is plain text instructions; it runs nothing and calls no service.
All reasons and checks

phileiny/h3-storyboard-skillAuthor:phileiny

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add skills/h3-storyboard

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Minimal accessThe tool runs nothing and makes no network calls.

  • The skill is plain text instructions; it runs nothing and calls no service.

This is third-party code. Review the repository files before installing.

What it does

h3-storyboard sits before you write a MiniMax H3 prompt. The skill gives a six-step order for breaking a script into shots, rules for how many emotional beats one shot can hold, the brow-eye-mouth order of expressions and a table mapping emotions to observable actions. The author tested the rules on their own clips. The headline finding: H3 silently drops facial instructions when one shot holds too many. A single seven-second close-up with nine beats produced a frozen face, while three shots of 2–3 seconds with one beat each landed every one. SOURCES.md sorts rules into verified, partly verified and inferred, and PSNR is proposed as a way to measure the result.

Who it is for. Directors and short-series creators working in 9:16 who generate video with MiniMax H3 through ComfyUI and lose emotion on screen.

Good fit when

  • A series script has to be cut into shots with one emotional beat each.
  • A character stands with a neutral face although the prompt describes shock.
  • An object comes out smaller than intended in a vertical 9:16 frame.
  • You want to check with a PSNR table whether an emotion happened in a clip.
  • A silent character must show feeling through the body, without a line.

Not a fit when

  • You need H3 prompt syntax, ComfyUI setup or quantisation; the neighbouring minimax-h3 skill covers that.
  • The video is made with another model, and the rules are tied to H3 behavior.
  • You want finished clips rather than a directing method.

Example request

Break my script into H3 shots: one emotional beat per shot, with expressions described from brows to mouth.

Limitations

The skill generates nothing and runs nothing: it is a set of instructions for an agent. Rules come from one project, a serialised 9:16 drama through Ref2VA and local ComfyUI at about 60 seconds an episode. Five rules have a controlled comparison, one is plausible and four are reasoned, including the beat ceiling per shot: 9 fails, 1 works, the middle is unmeasured. Part of the skill text is in Traditional Chinese. Results depend on the H3 model version.

How to disable. Delete the `h3-storyboard` folder from the project or home `.claude/skills/` directory.

FAQ

How do I install h3-storyboard in Claude Code?

Run `npx skills add https://github.com/phileiny/h3-storyboard-skill --skill h3-storyboard`. You can also copy `skills/h3-storyboard/` into `.claude/skills/` by hand.

Why does the character's face stay still?

H3 silently drops facial instructions when a shot holds too many. Split the shot so each one carries a single emotional beat.

Does it replace minimax-h3?

No. The minimax-h3 skill covers prompt syntax and ComfyUI, while h3-storyboard answers what to put in the prompt. They do not overlap.

How do I check that the emotion landed?

Compare frames by PSNR: a frozen face reads 37–42 dB, a played emotion around 22 dB. The table is inside the skill.

Are the rules verified?

Partly. SOURCES.md splits them into verified, partly verified and inferred, and five rules have a controlled comparison.

Discussion

No questions or reviews yet.

Skills

A director skill carries a short drama from script to a ready MiniMax H3 prompt package with H1 to H36 checks.

SkillExtended access65

Раскадровка в одном скилле: из сценария выходит сетка 3 на 3 с режиссерскими кадрами и таблица таймкодов по сценам.

SkillMinimal access241

The skill retells a reference video, waits for your confirmation and writes MiniMax H3 prompts per scene and shot.

SkillExtended access325

A skill that turns a plot or scene into a cinematic AI video prompt for 6 to 30 seconds, with shots, performance and sound.

SkillMinimal access148

A Codex plugin that turns a story or brief into a drama or commercial, from characters to editing through ComfyUI

PluginExtended access61

README in short

The repository describes a directing method for the H3 model and states which rules are verified and which are inferred. Installation uses `npx skills add` or manual copying.

README badge

Original SKILL.md

---
name: h3-storyboard
description: Turn a script into MiniMax H3 shot lists with working emotional performance — beat tables, shot counts, expression direction, and the acting techniques that actually render. Use when breaking a script or scene into H3 shots, writing beat timings, directing a character's face or emotion, fixing flat/wooden performance, deciding shot length or how many beats a shot can hold, or writing prompts for episodic narrative work. Complements the minimax-h3 skill, which covers prompt syntax and ComfyUI setup but not performance or shot breakdown.
---

# H3 分鏡與表演

`minimax-h3` 那份 skill 教你**怎麼把已經想好的東西寫成 H3 格式**。
這份教你**怎麼想**——劇本怎麼拆成鏡頭、情緒怎麼變成模型演得出來的動作。

官方和社群的 skill 都停在翻譯層，沒有人做這段（`instann/minimax-h3-director`
把「自動拆鏡」列在 roadmap，還沒做）。下面全部是實拍歸納出來的，
每一條都標了驗證狀態。

---

## 零之二、⭐⭐⭐ 寫每一顆之前：先問【這顆要呈現什麼】，再決定構圖，最後才寫內容

**構圖不是「內容寫完之後順手決定的東西」。它是第一個決定，因為它決定了其他東西演不演得出來。**

```
✗ 常見的寫法（也是最容易失敗的）
  內容 → 表演 → 對白 → ……最後順手填一個景別
  結果：景別永遠是最中性的那個（置中、平視、正面），
        而中性構圖 ＝ 訪談感、證件照感 —— 它不會傳達任何情緒強度

✅ 正確的順序
  ① 這一顆在戲裡要達成什麼？（一句話，用觀眾的感受寫，不是用動作寫）
     「讓人覺得這是他這輩子最高的一刻」而不是「他抬頭看大家」
  ② 什麼構圖能達成它？   ← 機位高度、景別、人在畫面裡的位置、前景、視線方向
  ③ 然後才寫表演、光、對白、道具
```

### 為什麼這個順序不能顛倒

```
表演   臉佔畫面 1/7 時，「下眼瞼微抬」無處可畫 → 先有景別，表情才有地方發生
特徵   體格、骨相在遠景會被抹平 → 先有景別，幾何描述才守得住
情緒   「高光時刻」「被排除在外」「他一個人」這些都是【構圖講的】，不是表情講的：
       仰角＝主導　俯角＝被壓　置中＝中性　偏置＋前方留空＝有去向

Report a problem