PNG To Editable PPT

When To Use

Use this skill when the request is:

• 单张图片转换为单张 PPT。
• 要求 PPT 可编辑，而不是整图贴入。
• 要保留整体版式、文字、模块框、箭头、流程图、图标。
• 图片中复杂照片类素材允许作为独立图片对象嵌入，保证内容不丢。

Core Rule

Do not use the source image as the only slide object. Rebuild the page with native PowerPoint objects:

• Text: editable text boxes.
• Module boxes/cards: editable rectangles or rounded rectangles.
• Arrows/lines/flowcharts: editable connectors and shapes.
• Simple icons: editable native shapes.
• Complex photos, screenshots, realistic objects: cropped picture objects.

Quick Workflow

1. Inspect the source image size and visual structure.
2. Create a JSON layout spec from layout_spec.example.json.
3. Use pixel coordinates from the source image. The script maps pixels to PPT EMUs 1:1 at 96 DPI.
4. Crop complex photo-like assets through the assets section.
5. Recreate all editable structure in elements.
6. Generate the PPT:

python png2ppt_skill/scripts/editable_png_to_ppt.py --spec path/to/spec.json

1. Verify output object counts and, when available, render-check with LibreOffice:

python png2ppt_skill/scripts/editable_png_to_ppt.py --spec path/to/spec.json --render-check

Files

• WORKFLOW.md: detailed repeatable process.
• SPEC_REFERENCE.md: JSON spec fields and supported element types.
• layout_spec.example.json: starter spec.
• scripts/editable_png_to_ppt.py: deterministic PPT generator.

Quality Bar

The generated PPT should contain many editable objects, not one pasted image. A good output normally has:

• text_shapes > 0
• picture_shapes only for complex raster assets
• shapes significantly larger than picture_shapes
• same slide aspect ratio as the source image
• no obvious clipped text or missing visual area after render-check