How I Created 1,200 Limited‑Edition WeChat Red‑Envelope Covers with AI Skills

Creating WeChat Red‑Envelope Covers with AI Skills

On a recent Spring Festival I generated four cover sets using the AI “Skills” model and published them on the WeChat Red‑Envelope Cover Open Platform.

Acquisition of the Skills

ByteDance Coze “WeChat Red‑Envelope Cover” skill (outputs watermarked image, requires Coze membership).

Self‑hosted implementation based on the NanoBananaPro image‑generation model, available at

https://skills.sh/wuchubuzai2018/expert-skills-hub/wechat-red-envelope-cover-designer

. The repository contains the skill definition and is still being refined (e.g., accessory handling).

Installation and Runtime Requirements

Install the skill with the same URL (click to add in Coze). After installation set the environment variable APIYI_API_KEY with a token obtained from the APIYI website; the skill uses this key to request an access token.

Generation Workflow

Open an AI development tool (e.g., OpenCode) and load the installed skill.

Provide a textual design request. The skill confirms the request, invokes NanoBananaPro to generate a full‑size image, then crops it to the dimensions required by WeChat’s review system.

Iterate through the following input fields until the preview is satisfactory:

Design theme (e.g., “Spring Festival, red, horse, personal symbols”).

Style keywords such as “minimal”, “premium”, “gold”, “whitespace”.

Layout proportion adjustments.

All steps are performed without Photoshop or prior design experience; the skill handles image creation and size adjustment automatically.

Design Concepts for Two Sample Covers

Both covers combine the following visual elements:

Dominant red background to evoke the holiday.

Traditional Spring Festival motifs (e.g., lanterns, fireworks).

A horse silhouette representing personal preference.

Gold accents for a premium feel.

The textual prompt supplied to the model explicitly listed these concepts; the model returned images that matched the description, after a few confirmation cycles.

Published Results

Four cover sets were uploaded; example images are shown below.

Effect 1:

Effect 2:

Effect 3 (the cover offered for claim):

Scarcity Rationale

The author limited each cover set to 1,200 copies, arguing that scarcity creates a participation effect: recipients are more likely to remember the sender, share screenshots, and inquire about the source.

Conclusion

The workflow demonstrates that an AI Skills capability can reduce the barrier to creating custom WeChat red‑envelope covers: a user supplies an idea, selects style keywords, and the skill produces a compliant image without manual graphic design.