Why AI images should have no text
Xiaohongshu covers often carry big titles, but having the AI draw Chinese text into the image is unreliable.
AI-rendered Chinese text often comes out with wrong characters, missing strokes or gibberish, so let the AI produce only the picture and add titles and captions in your layout tool.
What goes wrong
Image models draw text by what looks right, not what is right. Short English words sometimes work; Chinese characters have many strokes and often come out wrong or as character-like gibberish. One wrong character makes the image unusable.
A safer approach
- State "no text anywhere in the image" in the scene description
- Leave empty space in the composition for a title
- Add titles, captions and tags in a layout tool where font and position are under your control
What this tool does
Every image prompt asks for no text, letters, numbers or watermarks, with a clean composition and empty space. Titles and copy are delivered separately as text for your layout.
- Can the AI make a cover with the title on it?
- You can try, but Chinese text errors are common enough that rework gets expensive in batches. Keep picture and text separate.
Updated 2026-10-17