Summary

This project tackles the creation of visually rich presentations and the conversion of image-based slides into editable formats. It targets users within the Codex AI environment, offering an AI-driven solution to automate the generation of "luxury," high-density, complex-layout PPTs and their subsequent conversion to fully editable PPTX files, streamlining presentation workflows.

How It Works

Leveraging GPT's image generation and visual analysis, the system first uses GordenImagePPTGen to create image-format PPT slides (PNGs) and an image-based PPTX from a theme. Then, GordenImage2PPTX reconstructs these into an editable PPTX by separating and reassembling background, layout, icons, and text layers. An optional GordenSuperPPTSkill orchestrates this A-to-B process. This AI-driven approach enables sophisticated design and accurate, high-fidelity reconstruction.

Quick Start & Requirements

Installation involves copying self-contained skill directories (GordenImagePPTGen, GordenImage2PPTX, GordenSuperPPTSkill) into "${CODEX_HOME:-$HOME/.codex}/skills/" within Codex. Dependencies include Bash, Python 3.x, pip3 install python-pptx pillow numpy, and a Codex environment using GPT 5.5 ("medium" inference). Codex's built-in image generation backend is utilized. Setup is a file copy operation.

Highlighted Details

• Modular Skills: Three distinct, reusable skills: image PPT generation, image-to-editable PPTX conversion, and an end-to-end orchestrator.
• Layered Reconstruction: Conversion reconstructs editable PPTX into four layers: background, skeleton, icons, and text, ensuring detailed fidelity.
• AI-Powered Design: Employs GPT for generating "luxury," "high information density," and "complex layout" presentations.
• Visual Process: Intermediate image outputs allow observation of GPT's generation steps.

Maintenance & Community

Authored by Gorden Sun, with thanks to the LinuxDO community. A PPT Skill exchange group is available via WeChat (ID: duge360). No other specific maintenance or community links are provided.

Licensing & Compatibility

Licensed for commercial use, requiring attribution to the Github source or author (@Gorden Sun). Specifically designed for and compatible with the Codex platform.

Limitations & Caveats

Strictly dependent on the Codex environment and its specific GPT models/visual processing; not for standalone use or easy adaptation elsewhere. Image-to-editable PPTX conversion is computationally intensive and costly, consuming significant GPT Plus quota per image.