Xiaohongshu Image Workflow

⚠️ Core Principle: Propose first, generate later

Understand content → Design proposal (2-3 directions) → User selection → Generate → Preview confirmation → Upload

Uncle Hua Design Aesthetic Profile

Preferences

• Texture and warmth — Organic elements such as paper wrinkles, handwritten strokes, seals, tapes, etc.
• Handwritten/calligraphy fonts — Especially suitable for note-style, personal sharing scenarios
• Text fills the screen — Xiaohongshu is for mobile vertical screens, text should be large enough to be seen at a glance
• Core element enhancement — Key numbers/keywords as visual heroes (enlarge 3 times, change color, add decoration)
• Clear structure — Main title > Subtitle > List, distinct hierarchy
• Warm tones — Cream color, warm orange, warm gold perform well

Dislikes

• HTML screenshots — Too flat, like PPT templates, no soul (only fallback for precise data tables)
• Cyber neon/dark blue background — Dark blue backgrounds such as #0D1117 are aesthetic forbidden areas
• Signature/watermark — Cover images should not have "花生", "花叔", "@花生"
• Excessive white space — Better to be full than empty

Proven Good Styles

Core Parameters

--resolution 2K

3:4 portrait aspect ratio, 1080x1440 pixels

--viewport-size=1080,1440

Step 1: Understand the Content

• Theme: What is this content about?
• Core keywords: Which words/numbers need to be made into visual heroes?
• Mood/tone: Suspense? Dry content? Warm? Shocking?
• Number and type of images: Single cover / Carousel set / Infographic?

Step 2: Design Proposal ✅ Must wait for user selection

Proposal Format

### Direction A: [Style Name]
- Visual style: [One sentence describing the visual sense, e.g. "warm notebook paper + brush calligraphy large text + hand-drawn small icons"]
- Color: [Background color + main color + accent color]
- Copy layout: [Which text to enlarge as hero, which as subtitle, overall arrangement]
- Mood: [First impression of users after seeing it]

Proposal Principles

1. Each direction has clear differences (at least one of style, mood, color is completely different)
2. Mark recommendations (explain why a direction is recommended based on content characteristics)
3. Copy should be specific (not "enlarge the title", but "the number '28' is a 200px hero element, emphasized with orange")
4. No more than 3 directions (too many choices make selection harder)

Proposal Example

Direction A: Handwritten note style (Recommended)

• Visual style: Cream checkered paper background + brush calligraphy main title + hand-drawn technology small icons
• Color: Background #FDF6EC + main color #D97706 (warm orange) + accent circle lines
• Copy layout: "Alibaba C4 Building" and "A group of Cantonese people came" fill the upper half as hero, "AI Open Source Champion" is highlighted in orange as visual anchor, "Qianwen APP" seal at the bottom right corner
• Mood: Friendly, real, like a friend sharing inside information

Direction B: Dark gold reveal style

• Visual style: Dark frosted background + large gold text + badge decoration
• Color: Background #1A1A1A + main color #E2B714 (gold) + white auxiliary
• Copy layout: "Global AI Open Source Champion" giant gold text fills the screen, "Big Factory Inside Information" gold badge at the top, subtitle in white at the bottom
• Mood: Shocking, inside information, sense of weight

• Select one directly → Enter generation
• Request mixing/adjustment → Modify the direction and confirm again
• Not satisfied with any → Ask for preferences and re-propose

Step 3: Generate Images

Build Prompt

Create a [style] cover for a Xiaohongshu post. 3:4 portrait aspect ratio, 1080x1440 pixels, high quality rendering.

VISUAL STYLE: [Expand from the visual style description in the proposal]
COLOR PALETTE: [Specific color description]
TYPOGRAPHY: text fills most of the canvas, oversized bold typography, clear visual hierarchy.

TEXT TO RENDER:
- [Main title — hero element, visually dominant]
- [Subtitle]
- [Other text elements]

The word/number "[core keyword]" is visually dominant, 3x larger than other text, with decorative emphasis.

IMPORTANT: Do NOT include any personal signature, watermark, or author name like "花生" or "花叔".

[1-2 sentences describing the mood of the image]

• Large text →

  text fills most of the canvas, oversized bold typography

• Core enhancement →

  the word/number "XX" is visually dominant, 3x larger than other text, with decorative emphasis

• Handwritten style →

  handwritten style Chinese text / brush calligraphy lettering

• Paper texture →

  warm cream paper texture with subtle grid lines, notebook page feel

• Clear structure →

  clear visual hierarchy with distinct heading, subheading, and list levels

• No signature →

  Do NOT include any personal signature, watermark, or author name

Two generation paths (output both every time)

File organization specification (must comply)

Article directory/
├── article.md
└── [article abbreviation]-xiaohongshu-images/          ← Subfolder
    ├── A-note-style-AI.png
    ├── A1-note-style-HTML-warm.png
    ├── A1-note-style-HTML-warm.html
    ├── B-newspaper-style-AI.png
    └── ...

[direction letter][variant number]-[Chinese style name]-[path AI/HTML]-[variant description].png

• Direction letter: A/B/C (corresponding to the design proposal direction)
• No serial number for AI path:

  A-note-style-AI.png

• HTML variants with serial number:

  A1-note-style-HTML-warm.png

  ,

  A2-note-style-HTML-green.png

• Folder name uses article keywords:

  [keyword]-xiaohongshu-images/

AI generation command

export $(grep GEMINI_API_KEY ~/.claude/.env) && \
uv run /Users/alchain/Documents/写作/.claude/skills/xhs-image/scripts/generate_image.py \
  --prompt "[full prompt]" \
  --filename "[direction]-[style]-AI.png" \
  --resolution 2K

run_in_background=true

Step 4: Preview Confirmation

Browser preview (must do)

open

open "[image path 1]" "[image path 2]" "[image path 3]"

Inline preview

1. Is Chinese text rendered correctly?
2. Is the ratio 3:4 vertical?
3. Does the style match the selected direction?
4. Are there no signatures/watermarks?

Design review (must do)

• Each image: Comprehensive score + 1 core comment + 1 optimization direction
• Finally give a summary ranking table, mark recommendations
• Users can decide whether to adopt optimization suggestions by themselves

• Satisfied → Step 5 Upload
• Text error → Use HTML fallback to render this image
• Wrong style → Adjust prompt and regenerate
• Major direction change → Go back to Step 2 to re-propose

Step 5: Upload to Image Hosting

python3 /Users/alchain/Documents/写作/tools/upload_image.py "[image path]"

HTML Screenshot Path (used when AI text rendering fails, for precise data tables, or when users request comparison)

npx playwright screenshot "file:///path/to/card.html" output.png \
  --viewport-size=1080,1440 --wait-for-timeout=1000

• Canvas:

  width: 1080px; height: 1440px

• Font:

  font-family: "PingFang SC", "Hiragino Sans GB", "Microsoft YaHei", sans-serif

• Safe area: 80px top and bottom, 60px left and right

Quick Reference

Chinese text rendering limits (AI path)

• Main title ≤ 7 characters
• Subtitle ≤ 15 characters
• Body text per line ≤ 20 characters
• Must be verified one by one

Uncle Hua Technology Account Color Matching

Golden Rules

• Title is large, bold, eye-catching (occupies 30-50% of the screen)
• Core numbers/keywords are visually enhanced (enlarge, change color, add decoration)
• Cover has large amount of information → arouse curiosity
• Carousel images have unified style
• Vertical 3:4, make full use of screen space
• No signature/watermark

Related Skills

wechat-image

image-to-slides

Reference Files

• references/style-gallery.md

  — Complete style library and prompt template

• references/design-guidelines.md

  — Xiaohongshu platform design specifications

Produced by Uncle Hua | AI Native Coder · Independent Developer Official Account「花叔」| 300k+ followers | AI tools and efficiency improvement Representative works: 小猫补光灯 (Top 1 in AppStore paid list) · 《一本书玩转DeepSeek》