What does this skill do, and when should you use it?
PPT Master (v6.3.1) is an open-source MIT workflow skill by Hugo He that runs inside any agent-capable AI tool and converts PDFs, DOCX files, web pages, or pasted text into natively editable .pptx decks: real slide masters, native shapes and connectors, data-backed charts and tables, native transitions and animations (off by default), and audio narration synthesized from speaker notes — even narrated video. It can also distill reusable brand/layout templates from references and fill your existing .pptx with new content while preserving its design. The pipeline runs locally; apart from model calls, data stays on your machine with no platform lock-in. You need Python 3.10+ and an agent tool — no coding required.
- Reads source materials (PDF, DOCX, web pages, pasted text), confirms a design spec with you first, then plans pages and exports a native DrawingML deck to exports/
- Routes to one of several procedures: ordinary generation (Default/Quick), image-to-PPTX, beautify an existing PPTX, create reusable templates, or edit a native PPTX
- Produces native shapes (187 Office preset contours, Boolean merges, connectors, freeforms) and, on demand, data-backed native PowerPoint Chart/Table objects
- Synthesizes audio narration from speaker notes and can produce a narrated/self-running video; adds native transitions and opt-in entrance animations
- Accepts your own .pptx templates: fills new content while preserving the design, or generates real slide masters/layouts (p:sldMaster / p:sldLayout) via the structured route
- Sources images via image_gen.py (AI generation) or image_search.py (stock search — zero-config on Openverse/Wikimedia; Pexels/Pixabay with free keys)
- Consultants and finance professionals: hand an Agent a quarterly report PDF and get a native PPTX you keep refining in PowerPoint, not a flattened image
- Teachers and lecturers: generate courseware from lecture notes, auto-synthesize narration from speaker notes, and export a self-running teaching video
- Teams with brand guidelines: supply your own .pptx template or references so the skill distills a reusable brand/layout workspace, then fills future content without breaking the design
- Anyone with a well-designed but outdated .pptx: keep the original layout and swap in new text, table, and chart data
- People who want to add transitions, animations, and narration to a finished deck rather than rebuild it from scratch
- Anyone expecting a perfect finished deck in one click — the author states plainly it's a tool, not a wishing well; the model sets the ceiling and manual polish remains (cheaper models mean more polishing)
- Web-only users without an agent-capable AI tool and an unwillingness to install Python — the pipeline must run locally; there is no hosted web version of the project itself (aside from sponsor-hosted agents)
- Scenarios that require SmartArt — SmartArt is a deliberate, documented omission, not a gap
How do you install this skill?
- The document is very long; this review did not read it end to end, so deep edge-case handling is unverified
- Shell / CLI
- Network access
- Local filesystem
Python 3.10+pip dependencies (requirements.txt)Pandoc (only for legacy formats .doc/.odt/.rtf/.tex/.rst/.org/.typ)Optional AI image generation API key (e.g. OPENAI_API_KEY / GEMINI_API_KEY)Optional Pexels/Pixabay keys for stock image search
Two main install routes, both after installing Python 3.10+:
Git clone (recommended; can git pull updates anytime):
git clone https://github.com/hugohe3/ppt-master.git
cd ppt-master
pip install -r requirements.txtZIP download: Code → Download ZIP on GitHub, unzip, then run pip install -r requirements.txt; if too large, grab the skill-only ppt-master-skill-*.zip (~50 MB, fully functional without bundled examples) from Releases.
Claude Code plugin marketplace (skill files only, not the full repo):
npx skills add hugohe3/ppt-masteror inside Claude Code:
/plugin marketplace add hugohe3/ppt-master
/plugin install ppt-master@ppt-master
Image search/generation needs optional API keys in .env (Openverse/Wikimedia work with no configuration).
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Please create a PPT from projects/q3-report/sources/report.pdf
- Please turn the following into a PPT: [paste your content here...]
- Here's my existing .pptx template — fill it with the content from this new material, keep the design, and export only the pages I pick
- Add narration read from the speaker notes to this finished deck and export a self-running version
Open the project folder in your agent tool, then ask for what you need in the chat panel. Recommended flow:
# Place source files in projects/, then reference their path in chat
python3 skills/ppt-master/scripts/update_repo.py # later updates (Git installs only)Chat flow: give a file path or paste text → the AI confirms the design spec (template, aspect ratio, page count; ⛔ blocking gates require your confirmation) → pages are authored → export to exports/<name>_<timestamp>.pptx. By default charts and tables export as SVG-derived editable DrawingML shapes (cross-app visual consistency); pass --native-charts-and-tables to substitute PowerPoint-native Chart/Table objects (Edit Data support, but may render differently across apps). The skill routes deterministically via routing.md and runs an attribution integrity gate (attribution_guard.py); on Windows, use python if python3 is unavailable.
What are this skill's strengths and limitations?
- Real native PowerPoint output: masters/layouts, native shapes with working adjustment handles, data-backed charts and tables — all continue editable in PowerPoint
- Free and open source (MIT), no subscription; data processed locally with no platform lock-in — any agent-capable tool (Claude Code, Cursor, Codex CLI, etc.) can drive it
- Broad capability set: generate, image-to-PPTX, beautify, template creation, native editing, narration audio/video — each route with an explicit contract
- Strong documentation and examples: online flipping demos, multi-style example projects, FAQ, and a dedicated Windows installation guide
- Quality depends heavily on the underlying model — the author recommends Kimi K3/Claude with a large context window (~1M tokens) plus AI image generation; other models show a quality gap
- No automated test suite is mentioned; the README concedes rendering may vary across apps, especially on the --native-charts-and-tables route
- SmartArt is explicitly unsupported; the SVG pipeline has a single PPTX export route, and PowerPoint's manual Convert to Shape is outside the supported contract
- Windows needs extra setup steps (PATH, execution policy); legacy formats (.doc/.rtf etc.) require additionally installing Pandoc
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| PPT Master this page | 64 · Recommended | ★ 62k | 12d ago | MIT |
| PPTX Presentation Skill (Open Cowork) | 69 · Recommended | ★ 2.2k | 12d ago | MIT |
| deepPPT | 67 · Recommended | ★ 28 | 2mo ago | MIT |
| Gorden PPT Skill | 61 · Recommended | ★ 3.6k | 4mo ago | NOASSERTION |
| PPT Agent | 53 · Use with care | ★ 910 | 4mo ago | NOASSERTION |
Unlike generic AI slide-webpage platforms, PPT Master outputs a genuinely native .pptx you edit in PowerPoint (masters, native shapes, data-backed charts) and runs locally. The author also contributed to Microsoft's open-source ResearchStudio, which automates research dissemination from paper to talk video, poster, and blog — complementary to this skill (any document to native PPT), not a competitor.
How did FollowSkills review this skill?
Extremely long doc (100k+ chars); only the opening was read. The process is written with strict phase gates, but generation quality wasn't verified by actually running it.
Static read only, no execution to verify output quality.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
This review was completed under FSRS 1.0. Dimension scores are not converted; it is queued for re-review under FSRS 2.0.
See the full review method →