WeChat Sticker Generator

Problem it solves

A WeChat sticker album usually needs 24 character-consistent, expressive, transparent-background assets. Manually prompting an image model can produce inconsistent characters, flat facial actions, misaligned grid crops, and leftover backgrounds. This skill turns the workflow into an executable pipeline: collect character, style, scene, mode, and grid parameters, prepare a reference image, then generate album-ready assets.

How it works

The main flow is: detect API key → create a sandbox directory → produce or convert base_reference.png → write params.json → build prompts → batch draw → remove background and slice → generate WeChat banner, cover, icon, and upload_info.txt. It supports static and animated modes: static mode cuts independent PNGs, while animated mode composes a GIF from 9 or 16 frames. Action prompts should use exaggerated body pose + facial change + optional prop, avoiding weak labels like “smile.” Background removal can use dual, opencv, or rembg, depending on foreground color and text protection. Without an API key, the skill switches to a manual path: it handles parameters, prompts, receiving user-supplied images, and slicing outputs.

Limits

It is best for personal WeChat sticker albums where character consistency and batch delivery matter. Note that animated mode generates one GIF per run, so 24 stickers require 24 iterations. 16-grid mode is smoother but slower. API failures require filling missing subdirectories. The placeholder banner/cover in wechat_export must be redrawn before use.