What problem does it solve? Writing speaker notes from a .pptx usually relies only on text boxes, missing charts, tables, SmartArt, screenshots, and text embedded in images, which leads to inaccurate or invented narration. This Skill reads the real deck through structured extraction, slide rendering, region-scoped OCR, and vision review, then writes notes grounded in every visible element and injects the clean script back into PowerPoint. ## Core Features & Use Cases - Multi-source slide evidence: Combines python-pptx extraction, raw OOXML text, rendered slide PNGs, and image-region OCR into a per-slide visible-element inventory. - Pause-aware pacing model: Budgets words or Chinese characters per slide using realistic speaking rates (110 wpm English, 165 chars/min Chinese) with reserved time for transitions and pauses. - Complete deliverables: Produces a .pptx with clean notes injected, a full rehearsal document (.docx or .md), and an optional vision-review packet. - Use Case: A researcher has a 15-minute conference talk deck with dense result charts. The Skill inspects every chart axis and value, drafts a ~1,300-word script within budget, and delivers a ready-to-present PowerPoint with notes. ## Quick Start Use the ppt-speech-writer skill to write speaker notes for my deck 'thesis-defense.pptx' for a 15-minute English talk and inject them into the PowerPoint notes pane.
