What it does

Nature Image2PPT turns slide screenshots, scanned PDFs and image-only PPTX files back into object-level editable PowerPoint. Its bundled CLI drives the whole lifecycle: a doctor preflight that also picks the OCR route, prepare to open a run, per-page claiming and dispatch, semantic region decomposition and object routing, a hard gate on formula rendering, QA plus visual review that must leave evidence files, and a final rebuild with revalidation. State lives only in page_jobs.json and each page manifest, and every write is confined to its owner directory so nothing escapes the run.

Works with

Python 3.10 plus the bundled requirements; OCR goes through Baidu AI Studio PaddleOCR, and image generation uses either the built-in tool or an OpenAI-compatible endpoint.

Source and safety

OCR sends page images to the Baidu AI Studio PaddleOCR service and image generation posts prompts to an OpenAI-compatible images endpoint on a user-supplied key, flagged as justified. (This review was conducted when the skill was listed, not as ongoing monitoring. Sources are beyond our control — evaluate skills before using them.)