Chapter 6: Multimodal & Rendering — let the AI "see" images and use Markdown in chat
This chapter is about the "content" side of chat: ① images & media (multimodal); ② Markdown, code and formula rendering.
Reading time: about 8 minutes.
6.1 Images & media: let the AI "see"
- Send images to AI: upload or take photos, with crop, rotate, pen annotations and mosaic annotations (mark the key point, then ask);
- View AI's images: thumbnails, preview, save for images/videos returned by the AI (you can save to your device);
- Attachments reach the AI as multimodal content or extracted text — depends on whether your current model supports vision (if not, ask it to describe the text content instead).
Fun beginner idea: "Take a photo of the back of your router and ask the AI what the admin address on it says." (Requires a vision-capable model.)
Note
Messages can attach multimodal files like images, but the model you choose must support multimodal input, otherwise an error is returned. Document files (PDF, Word, etc.) are not supported yet.
6.2 Rendering abilities: tricks in chat
- Markdown: bold, lists, tables all render;
- Syntax highlighting: code blocks are auto-colored;
- Math formulas: LaTeX renders locally (e.g.
$E=mc^2$); - Inline images: image links/Data URLs from the AI preview directly.
6.3 Multimodal FAQ
Q: Why does the AI say "I can't see this image"? A: Your current model doesn't support vision input. Switch to a vision-capable model (GPT-4o, Gemini, Claude series, etc.) — see Chapter 8, 8.5.
Q: How do I save images output by the AI? A: Multimodal files returned by the AI support preview and saving — you can save them to your device.
End of chapter · Next: Personalization & Themes — dress up the app to your liking