Making a full WeChat sticker album that passed review with the open-source weixin-sticker skill: Codex handles generation, slicing, banner/cover/icon, checks and packaging
Chinese creator @ZhangZhangWanAI got several AI-made sticker albums approved and listed on WeChat's sticker platform. He found that "generate 16 cute stickers" is only the start; the real time sink is the submission package: 240×240 transparent singles, a banner, cover, chat icon and tipping images, then checking sizes and counts. He turned the process into an MIT-licensed skill that Codex, WorkBuddy and other agents with image generation can run. We ran its processing, validation and packaging scripts on our test machine.
Visit site weixin-sticker Skill (Codex / WorkBuddy + image generation) ↗- Pricing
- Free
- Region
- China
- Where it's from
- Chinese community
- Added
- Oct 21, 2026
- Tested on
- Oct 21, 2026
- Version we used
- zxd-666/weixin-sticker main @ 35a0149 (2026-10-20); scripts run with Python 3.13.5 + Pillow 12.3.0
- Original author
- 张口说AI @ZhangZhangWanAI ↗
- Our test
- Partly tested
- Cost / time
- The skill is free (MIT). Image generation costs whatever your agent's image model and plan cost.
The task
Turn yourself or an original character into a static sticker album ready to submit to WeChat's sticker platform (the stickers used in China's main chat app), without hand-slicing images, fixing sizes or making the extra assets yourself.
Steps
- Install: put the whole repo in your agent's skills folder (for Codex: git clone https://github.com/zxd-666/weixin-sticker.git ~/.codex/skills/weixin-sticker), then pip install -r requirements.txt (Pillow only). Copying just SKILL.md leaves out the slicing and checking tools. WorkBuddy and Doubao's agent import the full repo through their own "upload skill" entry.
- Start a new chat with the line below and attach a clear reference image of the character (for a real person, a photo of yourself or one you have rights to), plus any features that must be kept.
- Pick a mode: "step-by-step" shows you the multi-panel sticker sheet, then the banner, cover and icon, and waits for your OK at each stage; "no further confirmation" lets the agent slice, add tipping images, check and package on its own and hand you a ZIP.
- The agent makes one transparent multi-panel sheet (one emotion or action per panel, nothing crossing the cut lines) and gives each panel a 1–4 character Chinese theme name; no text in the artwork by default.
- It then makes a 750×400 opaque banner with no text, a 240×240 transparent cover and a 50×50 transparent chat icon (drawn separately as a front-facing head, not a shrunken cover); if you want tipping enabled, also a 750×560 prompt image and a 750×750 thank-you image.
- Slicing: the repo's offline page assets/表情包切割.html (grid or smart split) exports each sticker as a 240×240 PNG named after its theme, such as 开心.png.
- Check and package: python scripts/validate_pack.py <album dir> checks count, format, size, transparency and file size; python scripts/package_album.py <album dir> "name.zip" builds the ZIP.
- Submit: open sticker.weixin.qq.com in a desktop browser, register an artist account, create a sticker album, unzip the package and upload stickers, banner, cover and icon to their fields, preview at chat size, then submit for review.
Prompt
Use weixin-sticker to make 16 static WeChat stickers from the character reference image I've attached. Style: felt cartoon. Scenes include hello, got it, speechless, off work, good night. No step-by-step confirmation needed: finish the full asset set, check it and deliver it as a ZIP.
What the author says
On 2026-10-20 the author posted that his AI-made stickers had passed review and were listed on WeChat: Felt Zhangzhang, a beaver set, plus product manager, day trip and boxing themes (the README lists albums of 16, 20, 16 and 24 stickers, notes that the approval status is the author's own statement, and includes a screenshot of a WeChat album page). On 10-17 he wrote that generating 16 stickers is only the start; the real work is keeping the character consistent with distinct emotions, slicing 240×240 transparent singles, making the banner, cover, icon and tipping images, then checking sizes, transparency, names and counts. The skill chains all of that, and works in Codex, WorkBuddy and similar agents.
Our test
We did not run the image-generation step (the skill ships no image model) and did not submit anything to WeChat, so we can't verify the approval claim. We only ran the model-free processing, validation and packaging scripts on our test machine, using synthetic test images rather than real stickers.
What we found when we tried it:Tested 2026-10-21 (repo main, commit 35a0149, Python 3.13 + Pillow 12.3): with 8 synthetic 512×512 transparent circle images, 600×600 transparent cover/icon sources and a 1500×800 banner source, optimize_png.py produced stickers of about 9 KB each and a 1.9 KB icon; validate_pack.py printed "OK: album files passed measurable checks (0 warning(s))"; package_album.py built a ZIP with 8 stickers and 11 files, Chinese filenames intact. Negative tests: removing one sticker (7 left) gave "need 8–24 images, got 7"; adding hello.png failed the Chinese-name rule; an opaque white image was rejected as an icon at the processing step. We did not test the offline slicing page or image quality.
How to get in
The original posts and README are in Chinese; the GitHub repo is public. WeChat's sticker platform (sticker.weixin.qq.com) is in Chinese and artist registration is tied to WeChat; check its registration FAQ for identity and payout requirements before planning around tipping.
Who it suits
Creators, small brands and community owners who want themselves, a pet or an original character as WeChat stickers, and anyone already using Codex or WorkBuddy with image generation. For non-Chinese readers, it's also a clear example of turning a messy deliverable into an agent skill with machine checks.
Pitfalls
- The repo ships no image model: your agent must generate images, read and write local files, and have Python 3 with Pillow. Results across agents depend on those capabilities.
- Source images for the cover, icon and transparent stickers must already have transparent pixels: optimize_png.py does not remove backgrounds. In our test an opaque white image used as an icon failed with "icon input needs real transparent pixels; remove its background first".
- Sticker filenames must be 1–4 Chinese characters (hello.png was rejected by validate_pack.py in our test), and the count must be 8–24 (7 failed in our test).
- Machine checks cover count, format, size, transparency and file size only. Character consistency, distinct poses, white fringes, jagged edges and typos need your own eyes, ideally previewed at chat size.
- The ZIP is just for organizing: WeChat's back end doesn't take a ZIP, so unzip it and upload each file to its field.
- The size table is the author's 2026-10-17 summary of the platform's rules, which can change; follow the upload page at the time. Passing the scripts, or the author's own approvals, doesn't guarantee yours will pass.
- For real people, use your own photos or ones you have rights to; celebrities and existing IP characters raise copyright and likeness issues.
More in this category
- Turn a selfie into a transparent animated sticker setImageSeedream 5.5JimengHuashu skillSkill is free; image gen uses Jimeng credits◐ Partly testedChina
- Two minutes to give agents GPT stills and GIFsImageopenai-imageMCPGrok BotMCP free; images use OpenAI credits◐ Partly testedOverseas
- One sentence to batch WebP and trim edgesImageimage-processingClaude CodePillowSkill free; needs Claude Code + Pillow◐ Partly testedOverseas
- One photo plus a minimal paper-print twinImageCodexSkillImage genSkill free; needs Codex image generation◐ Partly testedChina
- Paint one dieline, fold a low-poly GLB with a Fold clipImageOrigamiClaude CodeBlenderSkill free; paint ~$0.10–0.25 each◐ Partly testedOverseas