What is gpt-image2-ppt-skills?

gpt-image2-ppt-skills is a Claude Code / Codex / OpenClaw skill that uses OpenAI's gpt-image-2 model to generate visually strong presentations from natural language prompts. It produces both an HTML viewer with keyboard navigation and a packaged .pptx file, and includes 10 built-in design styles plus template cloning capabilities.

How do I install gpt-image2-ppt-skills?

The recommended installation method is to paste a prompt into your AI assistant (Claude Code, OpenClaw, Codex, Cursor, Trae, or Hermes Agent) pointing to the install.md URL. Alternatively, you can clone the repository to ~/.claude/skills/gpt-image2-ppt-skills and run pip install -r requirements.txt.

Does gpt-image2-ppt-skills require an API key?

An OpenAI API key is required for the default openai backend. However, the skill also supports a codex CLI backend that reuses your existing Codex login, eliminating the need to configure a separate key within the skill itself.

What are the 10 built-in styles in gpt-image2-ppt-skills?

The skill includes Spatial Glass, Tech Blue, Editorial Mono, Dark Aurora, Riso, Japanese Wabi, Swiss Grid, Hand Sketch, Y2K Chrome, and Vector Illustration. Each style has three variants: cover, content, and data layouts.

How does template cloning work in gpt-image2-ppt-skills?

You can drop any .pptx file or image into the skill. It automatically renders the input, uses vision to extract the visual style, and applies a JSON Schema to replicate the layout with new content. This preserves the original's composition, color palette, and illustration vocabulary while swapping in your material.

gpt-image2-ppt-skills

by JuneYaooo · Python · Apache-2.0 · 697 stars · 31 forks

Open Source AI-Powered Agent Skill

Best for: Rapid visual PPT generation, template cloning, AI-assisted design workflows

Not for: Manual slide editing, offline use without API access, strict brand-guarded corporate templates

What It Does

gpt-image2-ppt-skills leverages OpenAI's gpt-image-2 model to transform natural language prompts into complete presentations. It operates as a skill within Claude Code, Codex, OpenClaw, and compatible agents.

🎨

10 Curated Design Systems

Spatial Glass, Tech Blue, Editorial Mono, Dark Aurora, Riso, Wabi, Swiss Grid, Hand Sketch, Y2K Chrome, Vector Illustration — each with cover/content/data variants

🪄

Template Cloning

Upload any .pptx or image; the skill extracts style via vision and replicates layout with new content through JSON Schema

🎮

Dual Output Format

Interactive HTML viewer (keyboard paging, auto-play, fullscreen, touch swipe) plus packaged 16:9 .pptx for direct use

⚡

Parallel Generation

Default 10 concurrent image requests; ~30 seconds for a 10-slide deck

🔁

Dual Backend

Direct OpenAI API (requires key) or Codex CLI backend (reuses existing Codex login, no separate key needed)

Installation

Recommended: Let your AI agent handle installation automatically.

            # Paste this prompt to Claude Code / Codex / OpenClaw / Cursor / Trae / Hermes:

            帮我安装 gpt-image2-ppt-skills：

            https://raw.githubusercontent.com/JuneYaooo/gpt-image2-ppt-skills/main/docs/install.md

Or install manually:

            
            git clone https://github.com/JuneYaooo/gpt-image2-ppt-skills.git ~/.claude/skills/gpt-image2-ppt-skills && cd ~/.claude/skills/gpt-image2-ppt-skills && pip install -r requirements.txt
        

Style Reference

gradient-glass

Apple Vision OS spatial glass

clean-tech-blue

Stripe / Linear blue-white

vector-illustration

Retro vector + black stroke

editorial-mono

Kinfolk / Monocle editorial

dark-aurora

Linear / Vercel dark neon

risograph

Riso dual-color + halftone

japanese-wabi

Muji / Kenya Hara wabi-sabi

swiss-grid

Bauhaus / Vignelli grid

hand-sketch

Sketchnote whiteboard style

y2k-chrome

Y2K liquid metal + stickers

Key Signals

GitHub Stars

697

Forks

License

Apache-2.0

Language

Python 3.8+

Last Updated

2026-05-23

API Key

Required (or Codex CLI)

Output Specs

Format: .pptx (16:9) + HTML viewer with keyboard navigation

Slide range: 5–20 slides typical

Concurrency: 10 parallel image generations by default

Model: OpenAI gpt-image-2 via official Images API

Base URL: Configurable for OpenAI-compatible proxies

Alternatives

Anthropic PPTX Skill

Official, free, no API key — but fewer visual styles

→

python-pptx

Programmatic control, no AI — requires manual design work

→

Community

Active development with recent commits. Topics tagged: ai-ppt, ai-presentation, claude-code-skills, openclaw-skills, pptx-templater.

style gallery

10-style cover demo

template clone

before / after

HTML viewer

keyboard navigation

.pptx export

16:9 packaged