JuneYaooo/gpt-image2-ppt-skills

gpt‑image2‑ppt‑skills – AI‑powered PPT creator

What it is

• A Python‑based skill that can be installed into multimodal AI agents (Claude Code, Codex, OpenClaw, Hermes, etc.).
• It uses OpenAI’s gpt‑image‑2 model to generate full‑page, high‑resolution slide images and then packages them into a standard .pptx file.
• Two output modes are supported:
  1. Visual‑first mode – each slide is a single rendered image (best aesthetic quality).
  2. Editable mode (off by default) – the generated slide is decomposed into native PowerPoint objects (text boxes, shapes, connectors, separate image layers) so the user can edit them directly.

Key capabilities

Feature	What you get
Template cloning	Upload any .pptx or image of a slide; the skill extracts its layout, colour palette and illustration style, then creates a new slide deck with completely different content while preserving the visual language.
Natural‑language slide creation	Ask the agent in plain English (or Chinese) for a PPT on a topic, specify length, style, or ask for a cover first, and the skill will generate the slides automatically.
Editable delivery	When the prompt includes “editable mode” the tool rebuilds each page into native PowerPoint objects (text, basic shapes, connectors, and separate picture layers).
Batch generation	Up to 10 slides are generated in parallel (≈30 s for a 10‑page deck).
Preview‑first workflow	By default the cover is generated for confirmation before the rest of the deck is produced.
Online template gallery	A web gallery (SlideCraft Template Gallery) lists dozens of pre‑validated templates; each entry provides a prompt that can be copied straight into the agent.
Versioning & rollback	The skill records which pages were regenerated and keeps previous versions, making iterative edits easy.
Real‑asset handling	User‑supplied logos, screenshots, charts, etc. are kept as separate image layers; they can be left untouched or merged into the AI‑generated design.

How to get it

1. Skill‑install via natural language – send the following to a supported agent:

   帮我安装 gpt-image2-ppt-skills：
   https://github.com/JuneYaooo/gpt-image2-ppt-skills
   

   The agent will copy the repository into its skill directory and restart.
2. Manual install (optional) – clone the repo and run the provided script:

   git clone https://github.com/JuneYaooo/gpt-image2-ppt-skills.git
   cd gpt-image2-ppt-skills
   bash install_as_skill.sh --target claude   # or codex, etc.
   

   The script places the skill under ~/.claude/skills/ or ~/.codex/skills/.
3. Configure API keys – set OPENAI_API_KEY (and optionally OPENAI_BASE_URL) in the environment or in the agent‑specific settings file. No keys are stored in the repo.
4. Ensure a PPTX rendering backend – Windows PowerPoint, macOS Keynote, or LibreOffice must be installed for the editable mode and for template‑cloning visual checks.

Typical usage patterns

• Create a brand‑new deck: “帮我做一份关于 AI 如何改变内容创作的 6 页 PPT，先生成封面确认，整体风格要有高级科技感”。
• Clone a corporate template: Upload company-template.pptx and ask for a new deck on “年度产品战略”。
• Edit an existing slide: “把第 3 页的副标题改成‘2026 目标’，删掉页脚”。 The skill regenerates only the affected page, preserving the rest.
• Get an editable file: “帮我生成一份可编辑模式的 PPT，文字和基础图形要能直接修改”。 The output includes ‑editable.pptx plus PNG renders for AI‑based verification.

Who benefits

• Business users needing fast, polished presentations (product launches, investor decks, training).
• Design teams that want to prototype many visual styles quickly.
• Developers of multimodal agents who need a ready‑made skill for PPT generation.
• Anyone who prefers natural‑language interaction over manual slide design.

License – Apache 2.0 (free for commercial and non‑commercial use).

---

All details above are taken directly from the repository’s README; no additional features have been inferred.