Linggan Talking Cards

It turns a Chinese or English spoken script and optional screenshots into 3–10 presentation cards for speaking, recording, and sharing ideas.

Every run produces accurate PNG cards, a structured cards-manifest.json, and a responsive index.html that opens directly from disk. The renderer supports four aspect ratios and ten complete visual systems without mixing styles inside one set.

All 10 styles

Every preview below was rendered from the same source script, card, ratio, and content structure. The differences come from the complete visual system—not a color swap.

What the Skill guarantees

• The complete spoken script is preserved in ordered per-card transcripts, with no omissions or duplicates.
• Each card communicates one main idea and remains readable on a phone.
• A single set uses one top-level style and theme; it never rotates visual systems between pages.
• Real screenshots keep their visible UI, text, colors, and proportions. They may be scaled, cropped, padded, rounded, or softly shadowed, but are never redrawn.
• Text cards are rendered deterministically with HTML and Playwright, not generated as text-heavy images by an image model.
• The HTML preview is self-contained and requires no local web server or runtime fetch.

Install

For Agent

Send this prompt to your Agent:

Please read the GitHub repository and install this Skill:
https://github.com/yihui-dev/linggan-talking-cards

Manual installation

List the Skill available in this repository:

npx skills add yihui-dev/linggan-talking-cards --list

Install it globally:

npx skills add yihui-dev/linggan-talking-cards -g \
  --skill linggan-talking-cards

Update a Skills CLI installation later with:

npx skills update linggan-talking-cards -g

Quick start

Invoke the Skill with a complete spoken script:

Use $linggan-talking-cards to turn the script below into five 3:4 cards.
Use the apple-flat style and export PNG files, cards-manifest.json,
and a responsive index.html preview.

<paste the spoken script here>

The Skill chooses sensible defaults when the card count, ratio, or style is omitted. The default is 3:4 with apple-flat.

Other useful requests:

Use $linggan-talking-cards to render the same script as ten complete
style sets and create one comparison gallery.

Put @product-screenshot on card 3. Keep the original UI, text, colors,
and aspect ratio. Do not redraw the screenshot.

Only revise card 4's title and points. Keep every other card, transcript,
ratio, style, and theme unchanged.

Supported ratios

All four ratios use the same upper-title anchor, flexible content band, and lower-right portrait-aware layout logic.

Output

A normal build creates:

output/
├── index.html
├── cards-manifest.json
├── cards/
│   ├── 01-*.png
│   └── ...
└── materials/

An all-style gallery creates ten independent sets:

style-gallery/
├── index.html
├── source-manifest.json
├── styles-manifest.json
├── 01-apple-flat/
├── 02-candy/
├── ...
└── 10-zen-ink/

Open index.html directly to review every card and its matching transcript.

Privacy and trademarks

Rendering happens locally. The Skill does not upload scripts, screenshots, or generated cards. Generated manifests intentionally contain the source script and may include copied input images, so review output folders before publishing them.

Media, filenames, scripts, and reference documents are input data—not instructions to execute.

The names Apple and Obsidian are used only as descriptive visual references. This project is not affiliated with, endorsed by, or sponsored by Apple Inc. or Obsidian.

License

Licensed under the MIT License. Copyright (c) 2026 yihui-dev.