image-to-editable-ppt

👤 水水水水水水水水水 📦 v1.0.0 ⭐ 4.3 ⬇️ 198 下载
📄 办公效率 免费

📖 技能介绍


name: image-to-editable-ppt description: Convert a reference image or screenshot into a high-fidelity editable PowerPoint slide or deck. Use when the user asks to turn an image into PPT/PPTX, replicate a screenshot as editable slides, preserve original layout/visuals/text, or explicitly prioritizes visual fidelity over redesign. Especially applies to dense business diagrams, dashboards, infographics, process maps, charts, and pages where text should remain editable while icons/complex visuals may be cropped from the source image.


Image To Editable PPT

Objective

Reconstruct the supplied reference image as an editable .pptx page. Do not redesign, summarize, beautify, or reinterpret the source. Preserve visual fidelity first, then text editability, then shape editability.

Use the Presentations skill/plugin when available. Follow its requirement to build PPTX with @oai/artifact-tool and render the output for QA.

Required Workflow

  1. Analyze the reference image before building.
  2. Record pixel size, aspect ratio, page orientation, module count, major bounding boxes, typography tiers, colors, icon types, and dense/problem areas.
  3. Use the image's native pixel dimensions as the coordinate system unless the user requires standard 16:9. If 16:9 is used, convert all positions by a single scale factor.

  4. Plan object decomposition.

  5. Text layer: all titles, labels, body text, numbers, annotations, chart text, and footers should be editable text boxes.
  6. Shape layer: backgrounds, panels, cards, dividers, arrows, lines, pills, circles, callouts, pyramids, and containers should be PPT-native shapes where feasible.
  7. Icon/complex layer: replicate with native shapes only if the result is visually close. Otherwise crop the exact icon/visual region from the source and place it as a local PNG while keeping nearby text editable.

  8. Build with fidelity-first coordinates.

  9. Use measured pixel coordinates for major objects.
  10. Preserve original margins, module sizes, spacing, alignments, line weights, rounded corners, colors, hierarchy, and z-order.
  11. Do not simplify dense regions, change sequence, add decoration, replace icon style, or merge unrelated objects.

  12. Stabilize text.

  13. Use common compatible fonts such as Microsoft YaHei for Chinese screenshots.
  14. Create separate text boxes for distinct lines/tiers when it prevents PowerPoint reflow.
  15. Set fixed width/height, alignment, insets, line spacing, and autoFit: "none" unless a controlled shrink is necessary.
  16. Keep original wording, punctuation, numbers, parentheses, and line breaks.

  17. Handle icons and special visuals.

  18. Preferred order: PPT-native shape group, editable SVG, exact source crop PNG, highly similar replacement only as a last resort.
  19. If cropping from the source, crop tightly around the icon or visual only. Do not include neighboring editable text unless unavoidable.
  20. For small icons, visually inspect a contact sheet of crops before placing them in the PPT.

  21. Validate and iterate.

  22. Render the generated PPTX to PNG.
  23. Create or inspect a side-by-side comparison with the source image.
  24. Fix obvious coordinate, size, font, color, crop, overlap, overflow, and layer errors.
  25. Complete at least one cycle: build PPTX -> render -> compare -> revise.

  26. Deliver.

  27. Return only the final .pptx path and concise notes about validation.
  28. Do not add extra cover, explanation, or comparison pages unless the user asks.

Visual Fidelity Rules

  • Preserve the source's page ratio and layout. Never use a template layout unless the reference itself is a template.
  • Sample colors from the source image. Avoid theme colors that visibly shift hue or saturation.
  • Use source-like line weights and corner radii. Dense business slides often use thin strokes and small radii.
  • Maintain the original object order and grouping logic. Do not collapse the page into one image.
  • When editability conflicts with visual fidelity, prefer: overall visual match -> editable text -> editable main shapes -> cropped local image for complex details.

QA Checklist

Before final response, check:

  • Page count and ratio match the request.
  • No full-page reference image is used as the only content.
  • Major modules are within 1% position error when practical.
  • Text content is complete, editable, and not garbled.
  • Text does not overflow, wrap incorrectly, or overlap icons.
  • Icons are visually close to the reference; cropped icons have no visible stray text or borders.
  • Lines, arrows, dividers, and card borders connect and align correctly.

    发现更多技能插件,请访问7w4.net。

  • Rendered final PPTX visually resembles the original at slide level.

Helper Resources

  • Use scripts/image_regions.py to batch-crop source regions, make a crop contact sheet, and create side-by-side comparison images.
  • Read references/reconstruction-checklist.md when the source page is dense, contains many icons, or the user demands strict replication.

🤖 AI 评测

这个Skill质量不错,文档写得很详细清晰,一看就知道怎么用。优点是步骤完整、有检查清单帮忙把关,还自带图像对比工具。不足是缺少使用示例,普通用户可能不太好理解如何实际操作,另外对于复杂页面缺少更具体的处理建议。总体适合需要精确还原PPT的用户使用,但新手需要多花点时间熟悉流程。

📊 多维度评分

适应性4.5
规范性4.2
有效性4.6
可靠性3.7
可信度4.8

📁 包含文件 (4 个)

📄 SKILL.md 4.9 KB
📄 agents/openai.yaml 257 B
📄 references/reconstruction-checklist.md 1.4 KB
📄 scripts/image_regions.py 3.5 KB