PPT Presenter — 带逐字稿的演讲级PPT生成器

为每一页生成完整演讲逐字稿 + 演讲者视图，让你自信上台不怯场。

Pipeline Overview

User Topic/Notes
     ↓
1. Content Planning (outline + page count)
     ↓
2. Per-Slide Content (title, board description, speaker notes)
     ↓
3. Image Generation (Gemini 3 Pro Image, one per slide)
     ↓
4. HTML Assembly (reveal.js + lightbox + presenter view)
     ↓
5. Speaker Scripts (word-for-word 逐字稿 in \x3Caside class="notes">)

Step 1: Content Planning

From user's topic, notes, or markdown files, produce an outline:

• Determine audience (technical level, mixed audiences)
• Plan 15-25 slides grouped into 4-7 sections
• Each section gets a divider slide (colored section number + title)
• Structure: Opening hook → Sections → Quick Start → Takeaway + Q&A

If user provides existing markdown/text content, read it and restructure into slide outline. Ask user to confirm page count and structure before proceeding.

Step 2: Per-Slide Content

For each slide, define:

• Page title — concise, with emoji prefix
• Visual description — what to draw on a whiteboard (colors, layout, diagrams)
• Key content — bullet points, tables, code blocks, comparison charts
• Speaker script — 150-300 words per slide, conversational tone, can be read aloud directly

For mixed audiences, write three-layer explanations:

• 编程小白 (beginners): metaphors, daily-life analogies
• 职场白领 (business): ROI, efficiency, practical scenarios
• 开发人员 (developers): architecture, code, protocols

Write content to a markdown file first for user review, then proceed to image generation.

Step 3: Image Generation

Generate one whiteboard-style illustration per slide using Gemini API.

Use the script: scripts/generate_slide_images.py

python3 scripts/generate_slide_images.py \
  --prompts-file prompts.json \
  --output-dir ./images \
  --api-key "$GEMINI_API_KEY"

Prompt Guidelines

• Style: "whiteboard marker drawing, hand-drawn educational style, colorful markers (red, blue, green, orange)"
• Include Chinese text in prompts when audience is Chinese
• Describe specific visual elements: diagrams, flowcharts, comparison tables, icons
• Keep prompts under 500 chars for best results

API Details

• Model: gemini-3-pro-image-preview (preferred) or imagen-4.0-generate-001
• Endpoint: https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent
• Request body for Gemini 3:

  {
    "contents": [{"parts": [{"text": "Generate this image: {prompt}"}]}],
    "generationConfig": {"responseModalities": ["IMAGE", "TEXT"]}
  }
  

• Response: candidates[0].content.parts[].inlineData.data (base64)
• Rate limit: add 2-3 second delay between requests
• Retry on network errors (SSL EOF, connection reset)

Finding API Key

Check TOOLS.md for Gemini Image Generation section, or ask user for API key.

Step 4: HTML Assembly

Use the template: assets/reveal-template.html

Structure

\x3C!DOCTYPE html>
\x3Chtml lang="zh-CN">
\x3Chead>
  \x3C!-- reveal.js 5.1.0 from CDN -->
  \x3Clink rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/reveal.css">
  \x3Clink rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/theme/black.css">
  \x3C!-- Custom styles (see template) -->
\x3C/head>
\x3Cbody>
\x3Cdiv class="reveal">\x3Cdiv class="slides">
  \x3C!-- Section dividers: colored background gradient + section number -->
  \x3Csection data-background-gradient="...">
    \x3Cdiv class="section-num">1\x3C/div>
    \x3Ch2>Section Title\x3C/h2>
  \x3C/section>
  
  \x3C!-- Content slides -->
  \x3Csection>
    \x3Ch2>Slide Title\x3C/h2>
    \x3C!-- Content: .two-col, .three-col, .card, table, .code-block, .timeline -->
    \x3Caside class="notes">Speaker script here...\x3C/aside>
  \x3C/section>
\x3C/div>\x3C/div>

\x3C!-- Lightbox overlay (see template) -->
\x3C!-- reveal.js init + lightbox JS (see template) -->
\x3C/body>
\x3C/html>

Key CSS Classes (defined in template)

Lightbox Feature

All images with .slide-img, .slide-img-large, .slide-img-full, or inside .photo-row are clickable. Clicking opens fullscreen lightbox with:

• Left/right navigation (arrows + keyboard)
• Counter (current/total)
• Close button + ESC key
• Pauses reveal.js keyboard while open

The lightbox JS is included in the template.

reveal.js Configuration

Reveal.initialize({
  hash: true,
  slideNumber: 'c/t',
  transition: 'slide',
  width: 1280,
  height: 720,
  margin: 0.06,
  plugins: [RevealHighlight, RevealNotes]
});

Keyboard hints shown at bottom: ← → 翻页 · S 演讲者视图 · F 全屏 · O 总览

Step 5: Speaker Scripts

Add \x3Caside class="notes"> to every content slide (not section dividers).

Script Guidelines

• Length: 150-300 words per slide (1-2 minutes speaking time)
• Tone: Conversational, as if speaking to audience directly
• Structure: Hook → explain visual → key points → transition to next
• Language: Match audience language (Chinese for Chinese audience)
• Include: Analogies, real examples, audience engagement ("大家看...")
• Avoid: Reading bullet points verbatim; scripts should expand on slide content

Presenter View

User presses S to open presenter view showing:

• Current slide (left)
• Speaker notes/script (right)
• Next slide preview
• Timer

Output Checklist

Before delivering, verify:

• All slides have titles with emoji prefixes
• All content slides have \x3Caside class="notes"> with full scripts
• All slide images generated and saved to images/ directory
• Images use .slide-img class (clickable lightbox)
• Section dividers have colored backgrounds and section numbers
• Lightbox JS is included and functional
• reveal.js loads from CDN (no local dependencies)
• File opens correctly in browser (open index.html)