Weibo posts & long-image format

Plenty of creators just copy their Xiaohongshu cover playbook to Weibo — and the metrics tank.

The reason? Weibo's visual logic is fundamentally different. It's a topic plaza, not a personal profile page. Your image isn't there to look pretty. It's there to make someone stop scrolling the topic feed and actually click your post.

Here's how the two platforms diverge:

This chapter gives you the full prompt templates for three Weibo cover formats (feed image / long-image / 9-grid), plus the real screw-ups I've hit.

---

1. Three Cover Formats

Format A: Feed Image (16:9, 1280×720)

Good for current-events commentary, news reposts, opinion pieces. Weibo users scrolling the feed are doing "small-screen quick-scan" mode — they need to grok the topic in half a second.

Format B: Long-Image (9:16 or custom, 1080×3000+)

A Weibo specialty. One image holds an entire article, so readers don't have to expand anything to finish it. Click-through rate beats text-with-link posts by 3-5x.

Format C: 9-Grid Photo Set (1:1 × 9)

Good for photography series, before/after comparisons, step-by-step tutorials. Weibo's 9-grid isn't just thumbnails — the platform stitches the 9 images back into one big image when users browse.

---

2. Feed Image Prompt Template

Horizontal feed image, 16:9, 1280×720.

Background: split screen composition.
- Left half: a screenshot mockup of ChatGPT interface showing a confused error message
- Right half: a real young person with shocked face reaction, dramatic lighting

Top center large red Chinese headline (huge bold, takes 1/4 height):
"ChatGPT 又更新了"

Bottom subtle bar text (white, smaller):
"这次他们做的是 ___"

Style: editorial news photography, dramatic split composition,
high contrast (red + black + white only), slightly gritty real-life feel.

Exact text only. No extra captions. Bottom 80px must be clean (microblog title overlays here).

Three Iron Rules for Feed Images

1. Type at 1/4 minimum. Weibo feed thumbnails are tiny. Anything smaller than 1/4 is unreadable.
2. Three-color palettes. Red + yellow + black, or blue + white + red. High-contrast trios are the safest bet.
3. Leave 80px clear at the bottom. Weibo overlays the username and timestamp down there. Pack it too tight and you'll get a collision.

---

3. Long-Image Prompt Template

The long-image is Weibo's killer move. One 9:16 image stuffs the key points of an entire article into a single visual — readers finish without ever clicking through.

Full template (good for a 3-minute-read article):

Vertical long-image, 9:16 ratio, 1080×3000.

[顶部 1/4 — 主题区]
Hero illustration: a brain with circuit overlay,
glowing neon blue + purple,
title "AI 时代如何选择副业" in huge bold white text on dark background.

[中部 1/2 — 内容区]
Four numbered cards stacked vertically, each card:
- Number badge (1, 2, 3, 4) on left side
- Bold Chinese category title on top of each card
- 2-line plain text body below the title
- Subtle divider line between cards

Card 1 title: "做技术变现"
Card 1 body: "Cursor / Claude Code 让一个人撑起一个产品"

Card 2 title: "做内容变现"
Card 2 body: "AI 写作 + AI 配图，单日产出 5 篇"

Card 3 title: "做服务变现"
Card 3 body: "AI 工具教学 + 1v1 咨询"

Card 4 title: "做信息差变现"
Card 4 body: "AI 工具评测 + 优惠码 / 联盟营销"

[底部 1/4 — 收尾区]
Author byline (small): "@匠人学院"
QR code placeholder (centered, 200×200, white border)
Footer text (very small): "长按识别关注"

Style: minimal infographic, dark navy + white + warm yellow accent.
Strict top-to-bottom flow, no diagonal elements.
Exact Chinese text only, all numbers as Arabic digits.

Three Principles for Long-Image Design

1. The top quarter needs a strong visual anchor. This is what decides whether readers swipe past your first screen.
2. Use card-based structure in the middle. Don't dump pure paragraphs — readers will burn out.
3. The bottom quarter is always the CTA. QR code, follow prompt, or tease the next post.

---

4. The 9-Grid

There are two ways to play the 9-grid on Weibo:

Approach A: 9 Standalone Images, Unified Tone

Good for photo series, monthly recaps, comparison tables.

Workflow: lock in image 1 with a full prompt, then use reference + short instructions to swap subjects for images 2-9.

Approach B: One Big Image Cut Into 9 Tiles

A 3240×3240 image sliced into nine 1080×1080 squares. Weibo auto-stitches it back into one big image when browsing — the effect is striking.

Square panoramic image, 1:1, 3240×3240 (will be split into 9 grid).

Subject: a panoramic view of a city skyline at sunset,
connecting nine emotional moments (each 1/9 panel tells a story):
- Top-left: morning coffee
- Top-center: subway commute
- Top-right: office arrival
- Middle-left: lunch with colleagues
- Middle-center: focus mode work
- Middle-right: afternoon meeting
- Bottom-left: evening gym
- Bottom-center: sunset commute home
- Bottom-right: night reading

Style: cinematic editorial photography,
warm to cool color gradient (top to bottom),
each panel distinct yet connecting to neighbors.

Slice it into nine 1080×1080 tiles in Photopea or Figma, then upload.

---

5. Real Screw-Ups

Screw-up 1: Long-Image First Screen Has No Hook

First time I made a long-image, I put "Foreword: in today's AI era..." at the top. Pure filler. Readers swiped past instantly, second-screen read rate basically zero.

Fix: The first screen of a long-image (top 1/4) needs a hook copy. "Why every big company is fighting over AI engineers" beats "Overview of the AI engineer role" by 10x.

Screw-up 2: 9-Grid Reading Order Got Scrambled

I arranged "steps 1-9" by visual aesthetics — diamond layout, diagonals — and users opening it found the reading order made no sense.

Fix: The 9-grid standard is left-to-right, top-to-bottom scanning. Anything sequential (steps, monthly recaps, Top 9 lists) has to follow that path.

Screw-up 3: Feed Thumbnails Get Crushed by Compression

Weibo's feed thumbnail compression is brutal. Upload at 1920×1080 and it comes out looking like mosaic tiles.

Fix: Output at 1280×720 directly (basically pre-compressed once) — it actually looks sharper on Weibo's end.

Screw-up 4: Forbidden Elements Get Blocked Outright

Mention Tiananmen, the national flag, or any well-known leader's name in your prompt and gpt-image-2 refuses. Even for ordinary news contexts.

Fix: Use generic descriptions instead. "A famous Chinese landmark with red architecture" instead of naming the actual landmark. "A national leader giving a speech" instead of a specific name. And dodge sensitive words when posting on Weibo too.

---

6. What We Learned

JR Academy's Weibo @JR匠人学院 ran a one-month long-image experiment with gpt-image-2:

What the numbers say: long-image reads are 22x pure text. Each long-image takes about 15 minutes (prompt tuning + post-production text proofing). Honestly the ROI is wild.

But long-images come with a "content barrier" — you need an 800-1500 word article first as the base, then compress it into the long-image. If you're just rehashing thin text, the effect actually goes the other way.

---

7. What's Next

Next chapter, Ch 12 Vertical Covers for Douyin / Channels / TikTok, covers the 9:16 cover playbook for the short-video era — a different logic from Weibo and Xiaohongshu.

If you're in a hurry to ship a long-image post:

1. Get an 800+ word article ready (or grab an old one)
2. Use the long-image prompt template from §3 to compress the key points into 4 cards
3. Sketch the structure on paper first (top / middle / bottom three sections)
4. Run the output through Photopea and check every Chinese character
5. Keep the caption under 140 characters (Weibo loves short copy + long-image)

Weibo's real traffic playbook is "long-image + one-line caption + repost-driven traffic" — this combo runs 3x faster than the WeChat MP equivalent.

---

📷 Real Weibo Post Examples

Three real Weibo post types, sourced from awesome-gpt-image (CC BY 4.0). Swap two variables and you can churn out weekly content at scale.

Example 1: Three-Day Travel Guide Cards (Series Content, Steady 10-20K Reads)

Prompt:

Generate a three-day travel guide image for [city]

The 3-day itinerary card is one of the classic Weibo long-image formats. Swap [city] for whatever city you're covering (Chengdu / Kyoto / Bali...) and you can crank out a series at scale. This kind of content holds steady at 10-20K reads — a reliable traffic source for weekly publishing.

📷 Creator: @MrLarus · Curated by: awesome-gpt-image

Example 2: Historical Figures Posting on Weibo (10K+ Recipe)

Prompt:

Generate a screenshot of [historical figure name] on [platform name]

Build a series account around "Li Bai posts on Weibo / Wang Xizhi posts on Douyin / Su Dongpo posts on Xiaohongshu". Pair it with captions that fuse historical stories and modern memes — repost rate runs 5-10x higher than pure text. This format crushes for culture creators across both Weibo and Xiaohongshu.

📷 Creator: @MrLarus · Curated by: awesome-gpt-image

Example 3: Encyclopedia-Style Knowledge Cards (Gold Mine for Science Accounts)

Prompt (excerpt of full version):

Generate a high-quality vertical encyclopedia-style infographic for [topic].
This should not be a normal poster but a modular educational infographic
combining a field guide, encyclopedia structure, lifestyle knowledge card,
and social-media explainer.
Include: clear main visual, several enlarged detail callouts,
multiple rounded modular information sections, strong title hierarchy,
visual scoring or Top 5 module.

The gold mine prompt for science and knowledge accounts. Outputs come with built-in modular cards, key-point tags, and Top 5 scoring blocks. Natural-history and science creators on Weibo can use this template to keep a steady output flow.

📷 Creator: @MrLarus · Curated by: awesome-gpt-image