Original video in Chinese.
Key Takeaway
- AI-native app: Google Slides uses the Nano Banana model to generate a complete PPT/infographic in one click, integrating images and text, far surpassing traditional “AI + old PPT” tools.
- Core advantages: content is form (infographic block structure, non-linear layout); contextual editing (natural-language image edits, operating in latent space rather than pixels); professional-grade output (such as a TPU vs GPU comparison chart), finishing in 1 minute work that would take hours by hand.
- Usage advice: first have an in-depth discussion of the content with Gemini in the sidebar, then copy it into Slides to generate for better detail; Chinese text currently suffers from crowding/missing strokes (a pixel-level generation issue), which the next generation of models may solve; investors should avoid similar projects, and users should just try Google Slides.
If you want to know what an AI-native app is, go try Google Slides. It’s Google’s version of AI PPT. But it’s completely different from those earlier AI PPT products.
Let me first show you what Google Slides looks like.
On the Slides page, click “New presentation,” and you’ll see a pop-up window. Let’s choose Slides, which is still in beta.
From the interface, Google Slides looks a lot like traditional PPT paired with an AI chatbot. Only this chatbot can directly generate images and text.
Let’s try one of the official presets, like the Industrial Revolution one. After waiting about a minute, a pretty solid PPT appears. If you have experience, you know how much effort it would take to manually make a PPT like this. And now, both the images and the text can be generated entirely with Google’s latest Nano Banana model.
Also, since it’s a PPT, frequent edits are inevitable. Click the edit button, and you can use the current image as context, then tell the AI in natural language what you want to change.
If you think the official preset example isn’t representative enough, let me show you how it performs in real use cases.
In the image generation feature, I asked it to design a PPT about the key design differences between TPU and GPU. From the result, it looks pretty professional, right?
However, if you want AI to make the entire PPT for you, I’d recommend using the infographic feature. I gave it the same request, and after about a minute, the infographic was done. Compared with the previous image, the content is basically the same, but I feel the infographic is more suitable.
So by now, you should understand what I meant at the beginning when I said Google Slides is more AI-native.
What do traditional AI PPT tools look like?
You enter a topic. AI goes off to do retrieval and matching. In other words, it finds an existing layout template, fills in the text, and then adds a not-very-relevant image.
Because earlier models were poor at multimodality, they didn’t understand the relationship between text content and image content, so those illustrations were purely decorative and couldn’t explain complex logic.
Those traditional AI PPT tools are essentially just “text filling machines.” Honestly, they’re pretty useless.
The logic of the new Google Slides is completely different. It has already achieved “pixel-level reasoning.”
The Nano Banana model does not just draw pictures; it also has logical reasoning ability. When you ask it to generate a PPT, it directly generates a brand-new image or infographic with data, based on an understanding of semantics.
With this capability, Google Slides can completely break the traditional structure of PPTs.
In the AI era, PPTs — if PPTs still need to exist at all — are no longer a pile of text boxes and illustrations, but are composed of high-density visual information blocks dynamically generated by AI. This is very much in line with the AI Native characteristic: content is form.
And you can also modify it through natural-language interaction, because it has contextual editing capabilities. The AI remembers the “layers” and “semantic structure” of the image. So what it modifies is not pixels, but the “latent space parameters” that generated the image.
If you’re an investor, don’t touch those AI PPT projects at home or abroad. If you can exit, get out now. If you’re a user, go try Google Slides. The only problem it has right now is that Chinese text generation still looks a bit ugly. Especially when there’s a lot of text, it gets all crammed together.
That’s because when Nano Banana generates infographics, the text is generated directly as image pixels, not rendered as vector fonts. The current model’s ability to do “pixel-level composition” for complex Chinese characters is still not as good as for English, so you get the phenomenon of “characters crammed together” or “missing strokes.” Let’s see whether the next generation of models can solve that.
Finally, one more thing. Let me share a very practical tip.
Google already has Gemini configured across its entire suite, including the Ask Gemini feature. So before generating anything, you can fully discuss it with Gemini in the sidebar first. Then paste the more detailed content generated by Gemini into Slides to generate from there. That way, the final result will be more detailed.
OK, that’s it for this episode. If you want to understand AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you next time!