YouMind
Sign in

Major Update: My 10-Year Design Experience Packed into PPT Skills

@op7418
SIMPLIFIED CHINESEMay 11, 2026
368K
1.7K
312
64
2.7K

TL;DR

The guizang-ppt-skill project receives a major update, adding a Swiss International Style, GPT-Image 2.0 integration for consistent visuals, and automated cover generation for social media platforms.

歸藏(guizang.ai) - inline image

After open-sourcing guizang-ppt-skill (github.com/op7418/guizang-ppt-skill) last time, everyone loved it, and in just a few weeks, it reached over 6,000 GitHub Stars. It was even referenced and included in the very popular open-source Claude Design.

I have also seen PPTs generated by this PPT Skill in many places, both online and offline.

歸藏(guizang.ai) - inline image

After the release, the most frequent questions I received in the background were:

"Can we have more styles?"

"Can you help me with the illustrations too?"

"Do I have to redraw the PPT cover for other platforms?"

I took notes while using it myself and saved them up for two weeks. This time, I'm filling all these gaps at once.

What's Updated

A brand-new visual style has been added, illustration capabilities are directly connected to Codex, and covers for platforms outside of PPT can now be made easily.

Specifically, three things:

  • Added Style B: Swiss Internationalism. Sans-serif throughout, single high-saturation anchor color, and grid-first.
  • Codex integrated with GPT-Image 2.0. Directly generate illustrations that match the tone, including film-textured portraits, flowcharts, and beautified UI screenshots.
  • Multi-platform cover generation. The same content can be used to create covers for Xiaohongshu, WeChat Official Accounts, Video Accounts, and more.
歸藏(guizang.ai) - inline image

Triggering the New Style

After installation, say to Claude or Codex: "Help me make a Swiss-style PPT."

If you have already installed it, you can also tell your AI: "Help me update guizang-ppt-skill."

It will ask you in return: International Klein Blue (IKB), Lemon Yellow, Lemon Green, or Safety Orange—which of the four themes do you choose?

  • International Klein Blue (IKB): Universal, business launches, AI products; default recommendation.
  • Lemon Yellow: Young, sports, retail, Y2K retro.
  • Lemon Green: Ecology, sustainability, Gen Z branding.
  • Safety Orange: Warning, news, high-energy themes.
歸藏(guizang.ai) - inline image

As before, custom hex codes are not accepted.

I specifically wrote this rule into the hard constraints of SKILL.md. I explained the reason in the previous article, so I won't repeat it.

Most Commonly Used Layouts

The Swiss style comes with 22 out-of-the-box named layouts, covering covers, chapters, data, comparisons, and closings.

Here are the six most recognizable ones:

Cover. The left half uses the IKB background with a line of giant white text, while the right half is white space with meta information. Almost a must-use for the first page.

歸藏(guizang.ai) - inline image

Statement. A single sentence occupies 9.6vw; the entire page is just that and a small line of footer text. Suitable for starting chapters or throwing out core arguments.

歸藏(guizang.ai) - inline image

KPI Tower. Four solid-colored columns with heights determined by data, followed by a row of categories. Longitudinal numerical comparisons like profit margin layering, price tiers, and conversion funnels can be understood at a glance.

歸藏(guizang.ai) - inline image

Loop Diagram. Numbered steps distributed on concentric rings, closing in a circle. Suitable for explaining self-learning loops, Agent automation cycles, and product feedback flywheels.

歸藏(guizang.ai) - inline image

Duo Compare. A hairline center line with a set of text and data on each side. Use it for old system vs. new solution, traditional vs. AI, or Before vs. After.

歸藏(guizang.ai) - inline image

Closing Manifesto. The left half is an IKB white-on-blue manifesto, and the right half has three takeaways to conclude the piece. Recommended for the last page of the deck to create a color loop with the cover.

歸藏(guizang.ai) - inline image

Others include horizontal timelines, Three Forces confrontation, System Diagram hierarchy, Why Now argument support, Tech Spec product specifications, Image Hero case studies with KPIs, and more.

Each corresponds to a typical content form. When you say "I want to do an industry ranking" or "I want to do a product benchmark," the AI will pick the most suitable one from these 22; you don't need to remember the names.

After choosing the theme, the rest of the process is the same as the previous version: 6 clarification questions, then the outline and rhythm table, and finally the code after alignment.

Let GPT-Image 2.0 Help You with Images

If you use it in the Codex environment, after the PPT is written, it will actively ask: "Do you want to generate some illustrations for this PPT?"

If you need them, it will automatically generate the appropriate type of illustration according to the selected PPT style and content:

Humanistic documentary photos (film texture), infographics (processes, comparisons, system relationships), screenshot redesigns (remaking your original images according to PPT proportions), data posters, flowcharts, and system relationship diagrams.

歸藏(guizang.ai) - inline image

Generated images will automatically adapt to your current deck's style and theme color.

Specifics:

  • For Digital Magazine style, infographics follow an e-ink tone: mainly black, white, and gray, with a few low-saturation accent colors, thin lines, grids, and restrained white space.
  • For Swiss style, infographics follow a Swiss modernism tone: sans-serif short labels with a Helvetica/Inter feel, 12/16 column grids, and right angles.
  • If you choose IKB Blue, the generated images will use IKB Blue as the only anchor color. The same applies if you choose Lemon Yellow, Lemon Green, or Safety Orange.
  • The text language follows the deck; Chinese decks use Chinese labels, and English decks use English.
  • Generated images do not have a PPT shell; they won't have headers, footers, page numbers, signatures, or decorative borders.
歸藏(guizang.ai) - inline image

This detail ensures you don't have to worry about the visual consistency of the entire deck.

You won't encounter a disconnect where the PPT is IKB Blue but the illustration is green, and you don't have to adjust prompts for GPT-Image while writing the PPT.

Visual drift across tools is the hidden cost I've struggled with most while using AI for content over the past two years. Solving it at the Skill layer means one less thing for the user to worry about.

Let It Help You Make Platform Covers

After the PPT is done, you always need to solve for three formats before posting: WeChat Official Account 21:9 header, Xiaohongshu 3:4 vertical image, and Video Account horizontal cover.

Just say: "Based on the core points of this PPT, give me a 3:4 Xiaohongshu cover," or "Give me a 21:9 WeChat header." The AI will generate images according to the Skill's visual rules (same theme color, same font, single focus rule).

歸藏(guizang.ai) - inline image

Batching is also possible. For Xiaohongshu's carousel mechanism, you can ask it to "batch produce 6 images with a unified style, consistent font size, and different layouts" all at once.

WeChat is special; the header is 21:9, but the sharing card uses a 1:1 square image, and the two need visual continuity.

The way to do it is to generate them twice but use the same set of colors and copy:

  • The first 21:9 main header has the main title on the left and a visual anchor on the right.
  • The second 1:1 square image serves as a visual extension of the sharing card, with the main elements echoing the header.
歸藏(guizang.ai) - inline image

My two commonly used layouts: solid background + giant text + corner date, or a split layout with an image on the left and text on the right.

A Few Clever Ideas Worth Sharing

Idea 1: Using film texture to combat the "AI feel"

When writing a share about "one person making a product," you need an image that conveys the emotion of "working alone." Previously, you'd spend half an hour on Unsplash or use an obvious stock photo. Now, just say: "Generate a 16:10 documentary photo of a person facing a screen in a late-night studio, natural light, low saturation, slight film grain, Fujifilm texture."

The resulting image has a restrained human warmth and doesn't look plastic. I wrote this judgment into references/image-prompts.md. Film texture is the greatest value of GPT-Image 2.0 for PPT illustrations; it removes the "AI-generated" label from the image.

歸藏(guizang.ai) - inline image

Idea 2: Redoing screenshots with weird proportions

Many creators have raw screenshots: product UIs, backend pages, data dashboards, all with different proportions and inconsistent white space. Putting them in a PPT ruins the visual flow. Now, just drop the original image to Codex: "Redo it in 16:10, keep all UI elements, add some real work scene environment, medium density."

It will regenerate a image that meets PPT standards, with all key information intact and perfectly unified proportions and white space. This is especially useful for product review content.

歸藏(guizang.ai) - inline image

Idea 3: "Wrapping" AI images with PPT templates

This is a usage I tested myself. If a GPT-Image 2.0 image is sent out alone, AI detection tools will likely flag it. This can affect recommendation weights on social media. But if you put that image into a PPT template and then take a screenshot of the whole thing, the result is completely different. This essentially assembles AI-generated material into a complete work, adding human traces. It's still an expression based on your outline's facts, not fiction, just avoiding being misidentified by algorithms.

歸藏(guizang.ai) - inline image

Why It Looks Good

Finally, let's talk about the design philosophy. Where did the visual rules for this Swiss style come from?

I gave it visual anchors like Massimo Vignelli and Helvetica Forever—the tradition of Swiss Internationalism from the New York Subway system, Unimark, and Müller-Brockmann.

If you've opened the Helvetica Forever book or seen the 1970 signage system Vignelli made for the NY subway, you'll recognize this language. Comparing the two styles, the difference is immediate:

歸藏(guizang.ai) - inline image

Simply put: Style A is for storytelling; Style B is for stating facts.

7 Design Disciplines

The visual rules of the Swiss style are actually just seven points, each an industry standard from a hundred years ago. I just wrote them into the Skill for the AI to execute.

1. Single Anchor Color. Only one highlight color is allowed in a deck. If it's blue, it's only blue. No blue-yellow collages. This is the soul of the Swiss style. "Less is more" is not just a slogan; it must be written as a hard rule in the code.

歸藏(guizang.ai) - inline image

2. Extreme Font Size Contrast. The ratio of the main title to the body text is at least 8:1. The giant manifesto on the cover uses min(11.6vw, 19vh), while the body is 1.1vw. Visual tension is pulled out by this contrast, not by decoration.

歸藏(guizang.ai) - inline image

3. The Larger the Text, the Thinner. Main titles use font-weight 200 (ExtraLight), not 700, 800, or 900. Swiss-style large text is like a scale on an architectural blueprint; it should be seen but not loud.

歸藏(guizang.ai) - inline image

4. Right Angles and Solid Colors. border-radius, box-shadow, and linear-gradient are all cut. All color blocks are right-angled, and all boundaries are 1px hairline lines. It looks harsh, but the "consumer app feel" or "SaaS template feel" you see everywhere comes from the abuse of these three attributes.

歸藏(guizang.ai) - inline image

5. Grid Supremacy. A 16-column grid with a 16px gap. All elements snap to the grid, using left alignment and large white space for asymmetric aesthetics. No centering, no even distribution.

歸藏(guizang.ai) - inline image

6. No WebGL Backgrounds. Style A uses WebGL fluids for a breathing feel; the Swiss style deliberately removes this. Pure white is its background; any dynamic background is a distraction.

歸藏(guizang.ai) - inline image

Closing Thoughts

The core of this update answers one question: how long is the chain of human-AI collaboration in content creation?

With the previous open-source release, I finished the "making PPT" part. This time, I connected "illustration generation" at the front and "multi-platform covers" at the back. Adding a new style and a new color loop closes the entire chain. From writing an outline to generating the PPT, illustrations, exporting, and publishing to different platforms—what used to require five software programs can now be done in one conversation.

Why does Master Guizang's PPT Skill still get so much usage and attention when there are so many PPT Skills on the market?

AI can only ever do 70% of the work. Every layout in these two templates was achieved through my manual fine-tuning on top of the AI. In other words, I am designing using natural language. Even in the AI era, 90-point content remains precious.

The Skill has been updated on GitHub: github.com/op7418/guizang-ppt-skill

The update method is the same as last time. The README has the "installation prompt for AI." Copy and paste it to your Claude Code, Codex, or any AI Agent with shell permissions, and it will automatically pull the latest version. Once installed, saying "Help me make a Swiss-style PPT" will trigger the new style.

One-click save

Use YouMind for AI deep reading of viral articles

Save the source, ask focused questions, summarize the argument, and turn a viral article into reusable notes in one AI workspace.

Explore YouMind
For creators

Turn your Markdown into a clean 𝕏 article

When you publish your own long-form writing, images, tables, and code blocks make 𝕏 formatting painful. YouMind turns a full Markdown draft into a clean, ready-to-post 𝕏 article.

Try Markdown to 𝕏

More patterns to decode

Recent viral articles

Explore more viral articles