Skip to main content
skill CX Codex

nanobanan-ppt — Claude Code skill: slide decks with AI images

nanobanan-ppt reads a document, splits it into slides and draws theme-matched images with Gemini. You get slide images and an HTML player, not a .pptx.

MIT 204 installs ★ 35 updated 2026-10-21

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add podo/design-agent-skills --skill nanobanan-ppt -a claude-code
Open on GitHub

What nanobanan-ppt does

The skill works through your source document for content and tone, then decides what belongs on each slide. Visuals come from Google's Gemini image model in one of two built-in styles: gradient glass or vector illustration.

It runs on Python 3 scripts, with generate_ppt.py as the entry point. Create a virtual environment, install google-genai, pillow, python-dotenv, PyJWT and requests, then copy .env.example to .env and put your GEMINI_API_KEY there. Image generation does nothing without that key.

Output is slide images plus an HTML player. Anyone who needs to fix wording in PowerPoint later is out of luck here. Transition videos between slides are optional and go through Kling AI, which asks for KLING_ACCESS_KEY, KLING_SECRET_KEY and FFmpeg on the machine. Upstream docs are written in Chinese.

What it helps with

  • turn a written document into a slide deck without retyping it
  • generate one matching set of images across the whole deck
  • page through the result in an HTML player
  • add video transitions between slides with Kling AI
  • choose a look: gradient glass or vector illustration

Example prompts for the agent

  • Make a deck from this document with NanoBanana, gradient glass style.
  • Build an AI PowerPoint: split the text into slides and generate an image for each one.
  • Run python3 generate_ppt.py --help in ~/.claude/skills/nanobanan-ppt and list the options.

How to install nanobanan-ppt

1Skills CLI (Vercel)

npx skills add podo/design-agent-skills --skill nanobanan-ppt -a claude-code

2Manually: put the files in place

~/.claude/skills/nanobanan-ppt/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/nanobanan-ppt/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about nanobanan-ppt

How do I install nanobanan-ppt in Claude Code?

Run npx skills add podo/design-agent-skills --skill nanobanan-ppt -a claude-code. The skill ends up at ~/.claude/skills/nanobanan-ppt/SKILL.md.

What does nanobanan-ppt do?

It analyzes a document, arranges the content into slides and generates styled images for them with Gemini. The result is slide images and an HTML player.

What do I need before generating images?

Python 3 and a GEMINI_API_KEY in the skill's .env. Video transitions also need the Kling AI keys and FFmpeg.

Source text

SKILL.md · MIT
---
name: nanobanan-ppt
description: |
  AI-powered PowerPoint generation with document analysis and styled image generation.
  Analyzes input documents, structures content into slides, and generates matching
  visual assets. Requires API credentials for image generation features.
triggers:
  - "NanoBanana"
  - "nanobanan ppt"
  - "AI PowerPoint"
  - "styled PPT"
  - "PPT with images"
  - "document to slides"
das:
  category: presentations
  upstream: "https://github.com/op7418/NanoBanana-PPT-Skills"
  upstream_path: "SKILL.md"
  version: latest
  install: false
---

# nanobanan-ppt

> Catalogue stub — full package: [op7418/NanoBanana-PPT-Skills](https://github.com/op7418/NanoBanana-PPT-Skills)

<!-- das:install:start — generated from stub.yaml by `npm run generate`; edit stub.yaml, not this block -->
## Decision tree

1. **Is the full skill already installed?**
   - `f=~/.claude/skills/ppt-generator/SKILL.md; [ -e "$f" ] && ! grep -qs "^das:" "$f" && echo "installed" || echo "pointer"`
   - `installed` → invoke it and proceed
   - `pointer` → go to step 2

2. **Install** (the skills CLI cannot install this one):

   Upstream's SKILL.md has no YAML frontmatter, so the skills CLI cannot see it; upstream's own README documents this clone into the Claude Code skills directory. The pip line adds PyJWT and requests, which upstream's video scripts import but its README omits.

   ```bash
   git clone https://github.com/op7418/NanoBanana-PPT-Skills.git ~/.claude/skills/ppt-generator
   cd ~/.claude/skills/ppt-generator && python3 -m venv venv && source venv/bin/activate
   pip install google-genai pillow python-dotenv PyJWT requests
   cp ~/.claude/skills/ppt-generator/.env.example ~/.claude/skills/ppt-generator/.env
   ```

**Heads-up**

- **Requires:** Python 3 and a Gemini API key (GEMINI_API_KEY in .env) for image generation; Optional, for the video feature: Kling AI keys (KLING_ACCESS_KEY, KLING_SECRET_KEY) and FFmpeg
- **License:** MIT per the README; no LICENSE file in the repository
<!-- das:install:end -->

## Invoke after install

First edit `.env` and add your `GEMINI_API_KEY` (video transitions also need the Kling keys), then check `python3 generate_ppt.py --help` from `~/.claude/skills/ppt-generator` with the virtual environment active. Image generation uses Google's Gemini image model; keep the key in `.env`.

- Skill name: `ppt-generator` (the directory it is cloned into; upstream's own SKILL.md has no frontmatter and calls itself ppt-generator-pro)
- Trigger phrases: "NanoBanana", "AI PowerPoint", "styled PPT", "document to slides"

## What it does

AI-powered presentation generation that goes beyond text structure: analyzes input documents for content and tone, structures content into coherent slide decks, and generates styled slide images matched to the presentation theme (upstream ships two built-in styles, a gradient glass look and a vector illustration look). Optionally generates transition videos between slides with Kling AI, plus an interactive player and a full-video export. Output is slide images and an HTML player rather than an editable .pptx file. Image generation requires external API credentials. The upstream docs are in Chinese.

## When NOT to use

- Slide decks that need no image-generation API → use `guizang-ppt` instead (it makes HTML decks and has no image-API dependency)
- When you don't have image API credentials configured
- When you need an editable PowerPoint file — use `anthropics-skills` (its `pptx` skill)

© podo · MIT · podo/design-agent-skills

Similar skills

All skills for Office documents →

From the same set108

The whole set →
wireframe-skill Turns a written screen description into a wireframe: a JSON structure plus one HTML preview where blocks can be dragged around. CX Codex +1 206installs swiftui-patterns swiftui-patterns is a Claude Code skill with idiomatic SwiftUI patterns for tab navigation, screen layouts, custom modifiers, and state handling. CX Codex +1 205installs data-analysis-skill The data-analysis-skill skill reviews Excel and CSV data, runs parallel expert analysis, and builds interactive HTML reports with charts. It can export the report to PowerPoint when requested. CX Codex +1 205installs threejs-claude-skill-package threejs-claude-skill-package — a bundle of 24 deterministic Three.js skills for scene building, WebGPU, React Three Fiber, physics, and architectural IFC models. CX Codex +1 205installs marp-slides marp-slides — a skill that walks the agent through Marp Markdown decks: slide density, visual hierarchy, plus ready SVG charts and a dark or light theme. CX Codex +1 204installs digidai-pm digidai-pm — a skill for Claude Code, Codex, Cursor, and other agents that adds senior PM thinking: SaaS health metrics, PRD critique, discovery, and PLG growth strategy. Targets B2B and consumer digital products. CX Codex +1 204installs data-viz-agent data-viz-agent builds charts that hold up: picks the chart type from the data relationship, orders encodings by perceptual accuracy, and checks contrast and keyboard access. CX Codex +1 204installs hig-doctor Points the HIG Doctor CLI at your interface code, whatever the framework, and turns the report into a severity-ranked list of Apple HIG and accessibility problems you can actually fix. CX Codex +1 203installs

Project sponsors

Companies that keep this analytics open