By Yijing Kai
The most annoying part of writing for WeChat Official Accounts isn't the writing itself—it's finding a cover image afterward.
Every time I finish an article and need to publish, I agonize over the cover. Before AI, I'd search Unsplash—which loads painfully slow these days. I even tried making them in Keynote, like this:
It's actually simple in concept—a specific image combined with title text. AI can totally handle this now, but letting AI generate freely tends to produce things that are "technically correct but weird." My content varies a lot in tone: some pieces are casual essays, some are tool reviews—dark moody vibes for certain topics, minimal abstract style for technical tools, magazine editorial feel for cultural commentary. Each one requires rethinking composition, color palette, and prompts. Quite tedious.
So I built a Skill to let Qwen Code + Model Studio CLI handle this for me.
Give it an article title or topic, and it automatically analyzes tone, picks a style, condenses the title, builds prompts, and calls Model Studio to generate the image.
Four styles covering my main WeChat content scenarios. Here's a quick look—still iterating on many aspects:
Image generation is powered by Alibaba Cloud Model Studio's command-line tool. Model Studio CLI is a multimodal AI toolbox in your terminal—chat, image generation, video generation, speech synthesis, knowledge retrieval, all from the command line. Here we mainly use image generate.
Why this approach:
--size '1344*572' outputs 2.35:1 widescreen ratio, Playwright @2x renders 2688×1144 high-res PNGBeyond that, Model Studio CLI can also:
If you're interested, check out Model Studio's open-source CLI directly: https://github.com/modelstudioai/cli
After installing my skill, just one prompt:
/open1v-mpcover-gen Generate a WeChat cover based on @your-article.md, <optionally specify style, e.g. use magazine style>Here's the full process for reference—you can also generate all styles at once:
❶ Install Qwen Code, see https://qwenlm.github.io/qwen-code-docs/zh/users/quickstart/
npm install -g @qwen-code/qwen-code@latest❷ Get a Model Studio API Key: https://bailian.console.aliyun.com/cn-beijing?tab=model#/api-key
[!important] Important
Go to the model usage page and enable "Stop when free quota is exhausted" to prevent charges: https://bailian.console.aliyun.com/cn-beijing?tab=model#/model-usage
❸ Install the cover skill. Launch Qwen Code and type:
Help me install this skill: https://github.com/joeytoday/open1v-mpcover-gen❹ Start using it. Just tell Qwen Code:
Make me a WeChat cover. The article title is "Qwen Code × Model Studio CLI Cover Generation Guide"
The AI will analyze content tone, ask which style you prefer (or auto-select), condense the title, build an English prompt, call Model Studio to generate the image, and output two alternatives.
Specify a style directly:
Make a minimal abstract cover about API security
Provide more context:
I wrote an article about Legalist philosophy. Make me a magazine-style cover.
Batch generation:
I wrote three articles this week with titles xxx / yyy / zzz. Make a cover for each.
❌ Not suitable for: precise Chinese typography (use Figma/Canva instead), real photographs, or highly refined cover designs.
Why not just have AI generate an image with the title text on it? Because AI-generated Chinese text is basically gibberish. So it's split into two steps: Model Studio generates the background → HTML overlays text → Playwright screenshots to export PNG. The image handles atmosphere only; text is precisely controlled by CSS.
The benefit is extremely low iteration cost—changing the title only requires editing text in HTML and re-rendering (2 seconds), no need to regenerate the image (30 seconds). Overlay too heavy or light, colors off—it's a one-line CSS change.
The four styles weren't arbitrary choices. They're divided by how readers feel in their content feed:
Each style maps to a text-image distribution pattern—full overlay, split columns, color overlay—ensuring text readability regardless of the background image.
The Skill is open-source on GitHub and continuously iterating. If you're using Qwen Code + Model Studio, give it a try. If you like it, a star ⭐️ would be appreciated.
🔗 https://github.com/joeytoday/open1v-mpcover-gen
🔗 https://github.com/modelstudioai/cli
🔗 https://github.com/QwenLM/qwen-code
[!note] Disclaimer
This content was written by the author, with AI-assisted editing and optimization. Core opinions and personal experiences are the author's own and represent only the creator's personal views.
Sharing Data Without Risking Leaks? Let Ticket-Based Enhanced Embedding Strike the Perfect Balance
1,415 posts | 496 followers
FollowAlibaba Cloud Indonesia - March 31, 2026
Alibaba Cloud Community - September 6, 2024
Kidd Ip - August 12, 2025
JwdShah - December 16, 2024
Kalpesh Parmar - May 12, 2026
Alibaba Cloud Big Data and AI - May 27, 2026
1,415 posts | 496 followers
Follow
Alibaba Cloud Model Studio
A one-stop generative AI platform to build intelligent applications that understand your business, based on Qwen model series such as Qwen-Max and other popular models
Learn More
Qwen
Full-range, open-source, multimodal, and multi-functional
Learn More
AI Acceleration Solution
Accelerate AI-driven business and AI model training and inference with Alibaba Cloud GPU technology
Learn More
Network Intelligence Service
Self-service network O&M service that features network status visualization and intelligent diagnostics capabilities
Learn MoreMore Posts by Alibaba Cloud Community