imagegen-scene-ppt

Generates image-based PPT decks and rebuilds image slides into editable PPTX files.

1|Updated Jul 13, 2026
One-click install
npx skills add https://github.com/HalseyYang/Skills-HY --skill imagegen-scene-ppt-halseyyang
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: imagegen-scene-ppt
Source: https://github.com/HalseyYang/Skills-HY/tree/main/codex-ppt-skill
Command: npx skills add https://github.com/HalseyYang/Skills-HY --skill imagegen-scene-ppt-halseyyang

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires Pillow, python-pptx, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Creating Chinese B2B business presentations forces a trade-off between visual quality and editability: AI-generated slides look polished but are locked inside images, while manually rebuilding slides in PowerPoint is slow. This Skill routes each request to the right path—full-page image generation for new decks, element-level reconstruction for editable PPTX, or SVG decomposition for lightweight reuse. ## Core Features & Use Cases - Full-page image deck generation: Writes per-slide image prompts, generates 16:9 pages via imagegen, overlays real logos in post-processing, and packages everything into PPTX/PDF with contact sheets. - Element reconstruction to editable PPTX: Converts existing image pages or screenshots into PowerPoint files with native text boxes, shapes, and transparent image assets, validated by diff heatmaps and layout QA reports. - SVG decomposition: Rewrites image pages as SVG structures for web or document reuse when object-level PowerPoint editing is not required. - Use Case: You have a screenshot-based sales deck that must be edited in PowerPoint. The Skill builds a visual inventory, generates isolated transparent assets, rebuilds the deck with native objects, and outputs a validation report confirming no full-page images remain. ## Quick Start Use the imagegen-scene-ppt skill to turn these slide screenshots into an editable PPTX using element reconstruction, with a validation report and render comparison.

Frequently Asked Questions about imagegen-scene-ppt

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I convert image slides into an editable PPTX?

Use the element reconstruction path: build a visual_inventory.json describing text, containers, and icons, generate transparent assets with imagegen, then run build_semantic_deck.py to assemble native PowerPoint objects. Validate the result with compare_render.py and validate_semantic_deck.py.

How to package generated slide images into a PowerPoint file?

Run scripts/package_image_deck.py or make pack with an images directory and output PPTX path. It crops images to 16:9, optionally overlays real logos, exports PDF via LibreOffice, and creates a contact sheet preview.

What is the difference between element reconstruction and SVG decomposition?

Element reconstruction rebuilds slides as native PowerPoint text boxes, shapes, and transparent image assets for real editability. SVG decomposition is faster and cheaper but produces SVG structures that PowerPoint imports as single graphics without stable object-level editing.

Can AI image generation create logos and QR codes for slides?

No. The Skill explicitly forbids generating real logos, QR codes, seals, or brand marks with imagegen. These must be overlaid in post-processing using package_image_deck.py's logo options to guarantee authenticity.

Why does my reconstructed PPTX have text overflow or overlapping text?

Text overflow usually comes from bbox or font settings in visual_inventory.json. Adjust bbox, role, parent_id, or fit_mode, or tune font policy in layout_rules.json, then rebuild and check the layout_qa section of build_report.json.

What are the limitations of converting images to editable PowerPoint?

Reconstruction is a rebuild, not decompilation, so complex pages reach roughly 75-90% visual fidelity. Cross-platform rendering differs between PowerPoint, LibreOffice, WPS, and Keynote, and missing fonts cause Chinese text wrapping drift.