Humanize PPT addresses the disconnect between visually appealing HTML presentation templates and the need for structured, engaging live talks. It targets technical users by transforming static decks into讲able (speakable) presentations using an AST-based narrative, integrated media, and automated quality checks, enhancing audience comprehension and engagement.

How It Works

The core is an AST (Audience-State-Transfer) approach, modeling audience comprehension evolution slide-by-slide to craft a narrative flow. It intelligently decides on incorporating real images, SVG charts, or Remotion videos, delegating generation to specialized downstream skills. A crucial "Speech Check" automatically identifies and flags pages unsuitable for live delivery, providing actionable feedback for revision.

Quick Start & Requirements

• Primary install / run command: npx skills add LearnPrompt/humanize-ppt -g Claude Code or via plugin marketplace.
• Non-default prerequisites and dependencies: Integrates with downstream rendering (guizang-ppt-skill, frontend-slides) and media generation (baoyu-image-gen via local Codex CLI, Remotion). baoyu-image-gen requires a ChatGPT subscription but no OpenAI API key.
• Links: Official docs are available within the repository.

Highlighted Details

• AST-Driven Narrative: Structures presentations around advancing audience understanding.
• Dynamic Media Integration: Automatically determines and assigns generation of images, SVG charts, and Remotion videos.
• Automated Speech Check (演讲体检): Scans rendered decks for presentation flaws (e.g., obscured text, incomplete narrative) and generates fix prompts.
• Style Gallery: Enables selection from multiple cover page designs before full rendering.
• Presenter Mode: Generates a presenter-shell.html with speaker notes and cues, usable before final deck rendering.

Maintenance & Community

Developed by "LearnPrompt". Community links include X (@aiwarts) and learnprompt.pro.

Licensing & Compatibility

Licensed under MIT, offering broad compatibility for commercial use. Emphasizes local execution with zero API keys.

Limitations & Caveats

Relies on compatible downstream skills for final HTML rendering. Automated Speech Check has limitations; some visual issues require manual review, and it caps at 3 QA iterations. Media generation requires local Codex CLI setup and a ChatGPT subscription.