← powerpoint.md

PPTAgent v2

Research-Heavy General Purpose Open Source Advanced Setup
Best for: Research-heavy presentations needing automatically generated visuals, deep topic coverage, and offline operation
Not ideal for: Quick one-prompt decks — setup complexity (ease 2.6/5) means this is a power-user tool, not a starter skill

Core Features

20+
Specialized tools
4.5
Output quality
T2I
Text-to-image
Offline mode
🔬
Deep Research Pipeline
Automatically researches your topic across multiple sources before generating slides — not just summarizing what you provide, but actively gathering supporting evidence, data, and context.
🖼️
Text-to-Image Generation
Generates slide visuals automatically. Diagrams, charts, and illustrative images are created and embedded — no stock photo hunting required.
🔧
20+ Specialized Tools
A full agent toolchain: web search, data extraction, chart rendering, layout selection, content refinement, citation formatting, and more — all orchestrated automatically.
📡
Offline Mode
Runs fully locally without sending data to external APIs. Critical for presentations involving confidential company or research data.

When PPTAgent v2 Is the Right Choice

⚠️ Setup complexity: ease 2.6 / 5 PPTAgent v2 requires Python environment setup and dependency installation. If you need a deck in under 5 minutes, use Anthropic PPTX Skill instead.
✓ Use PPTAgent v2 when your topic needs substantive research (market analysis, scientific briefings, policy papers, competitive landscapes)
✓ Use PPTAgent v2 when you need auto-generated charts and diagrams embedded directly in slides
✓ Use PPTAgent v2 when data privacy requires offline processing
✗ Skip PPTAgent v2 when you just need a clean, well-structured deck without deep research

Signals

Publisher
ICIP-CAS (research lab)
Maintenance
Actively maintained
Install Method
uvx (Python)
License
Open Source
Output Format
.pptx
Our Test Score
4.5 / 5

Quality Assessment

Output Quality
4.5
Ease of Use
2.6
Research Depth
4.8
Visual Generation
4.4

Install verified: pip install succeeds but pulls in 144 packages (playwright, opencv, modelscope, gradio, etc.) — the heaviest dependency footprint of all skills tested. After install, requires interactive pptagent onboard to configure LLM API key, then a template PPTX before generation is possible. End-to-end generation not independently tested due to API key requirements, but code inspection confirms a sophisticated multi-phase pipeline (Document → Outline → Slide Agent with LLM+Vision). Ease of use score reflects confirmed setup complexity: plan 30+ min for first run.

Install & Use

# Install via uvx (recommended, no manual clone needed) uvx pptagent generate "Your presentation topic" -o output.pptx # Or install with pip: pip install pptagent pptagent generate "Your topic" -o output.pptx # First-run auto-installs on macOS: Docker, Node.js, Playwright, Poppler # Linux users: install these manually before first run # Offline mode (no external API calls): pptagent generate "Your topic" -o output.pptx --offline # Specify slide count and language: pptagent generate "AI in healthcare 2026" -o deck.pptx --slides 20 --lang en # Example prompts that work well with PPTAgent: # "Global EV market analysis 2025 with competitive landscape" # "Quantum computing: state of research and commercial applications" # "Climate policy comparison: EU vs US vs China"

Alternatives & Tradeoffs

Anthropic PPTX Skill
Far easier to install and use (ease 4.8/5), best for quick clean decks without deep research
MiniMax PPTX Generator
Better design consistency via style recipes; no auto-research or image generation
PPT Master
Better for document-to-PPTX conversion; PPTAgent v2 wins when you're generating from a topic, not a source file
Financial Services Plugins
If your research is financial data specifically, that skill has direct FactSet/PitchBook integrations PPTAgent lacks

Community Outputs

Research-heavy decks generated with PPTAgent v2. Submit yours after use.
EV Market 2025
Claude 4 · Research · 4.6/5
Quantum Computing
Claude 4 · Science · 4.5/5
Climate Policy
Claude 4 · Policy · 4.3/5
AI Market Report
Claude 4 · Tech · 4.5/5