t4ku Speech Slides Skill
Generate slide decks from speech scripts in Projects/syla/syla-monday-morning/.
Workflow
- Read the speech script (## スピーチ原稿 section)
- Select template pattern (or let user choose)
- Generate PPTX using PptxGenJS directly (see references/pptxgenjs-guide.md)
- For charts: generate with D3.js, embed as SVG/PNG (see references/d3-charts.md)
- For custom images: use AI image generation skill (separate skill)
- Output PPTX to the speech's directory
IMPORTANT: Use PptxGenJS directly
Do NOT use the pptx skill's html2pptx.js — it has known silent failure issues with environment detection. Instead, use PptxGenJS directly:
# Ensure pptxgenjs is installed locally in the output directory
cd <output-dir> && npm install pptxgenjs
Write a .mjs script that imports PptxGenJS and calls pres.writeFile() with explicit Promise handling. See references/pptxgenjs-guide.md for patterns and helpers.
Template Patterns
4 patterns available. See references/templates.md for slide-by-slide structure.
| Pattern | Slides | Best For |
|---|---|---|
| A: Story Arc | 5-6枚 | 物語型スピーチ（導入→展開→結末） |
| B: Framework | 4-5枚 | フレームワーク紹介型（3つのポイント etc.） |
| C: Contrast | 4-5枚 | 対比型（Before/After、成功/失敗） |
| D: Minimal | 3枚 | インパクト重視（問い→核心→締め） |
Design System
const DESIGN = {
colors: {
CREAM: "F5F0E8", // background
DARK: "2B2B2B", // title/closing slides, callout boxes
AMBER: "D4882B", // accent, highlights, table headers
BLK: "1A1A1A", // primary text
WHT: "FFFFFF", // card backgrounds
RED: "cc3333", // negative/failure
GRN: "33aa55", // positive/success
GRY: "999999", // secondary text
TEAL: "4DD4D0" // sparingly, for special highlights
},
fonts: {
heading: { face: "Arial", size: 36, bold: true },
subhead: { face: "Arial", size: 24 },
body: { face: "Arial", size: 18 },
caption: { face: "Arial", size: 13 },
quote: { face: "Arial", size: 18, italic: true }
},
layout: "LAYOUT_16x9", // 10" x 5.63"
spacing: { margin: 0.8, gap: 0.3 }
};
Slide Types
| Type | Description | PptxGenJS Pattern |
|---|---|---|
title |
Big text + subtitle, dark bg | addText centered, bg: DARK |
statement |
Large bold statement | addText with accent color |
two-column |
Left text + right visual | Two addText/addShape at x=0.8 and x=5.4 |
comparison |
Side-by-side cards with color-coded top border | addShape(ROUNDED_RECTANGLE) + addShape(RECTANGLE) for top bar |
framework |
3 cards in a row | Loop with x = 0.8 + i * 3.1 |
table |
Comparison table | addTable with header fill |
quote-card |
Large quote, dark bg | addText italic, bg: DARK |
data-point |
Big number + context | Large fontSize number + smaller context |
closing |
Memorable line, dark bg | addText centered, bg: DARK |
Speech-to-Slides Extraction Rules
- Extract 1 key message per slide (max 20 words for heading)
- Use speaker's own words for headings — don't rephrase
- Include supporting data/quotes as body text
- Target: 1 slide per minute (5 slides for 5 min)
- Slides SUPPORT the speaker, don't replace them
- 60-30-10 color rule: 60% background, 30% text, 10% accent
Visual Enhancement Options
Charts (D3.js)
For data-driven slides, generate SVG charts with D3 and embed as images. See references/d3-charts.md.
AI Images (separate skill)
For custom illustrations/backgrounds, use AI image generation. Prompt patterns:
- Hero:
"{theme} hero image, professional, 16:9, high quality" - Icon:
"minimal icon for {concept}, flat design, transparent background" - Background:
"{theme} abstract background, subtle, suitable for text overlay"