GZH Cover Studio

The Problem It Solves

Creating WeChat Official Account ("GZH") covers often involves tight deadlines, the need for multiple aspect ratios, and persistent issues with AI-generated Chinese text (garbled characters, errors, or blurriness). This skill addresses these by separating visual generation from typographic composition, delivering high-quality, click-worthy covers quickly. It focuses on solving: rapid iteration for busy content teams, cross-platform sizing (main cover, thumbnail, Moments), and reliable Chinese title readability.

How It Works

The core methodology is a "background-first, text-overlay" pipeline, with key steps:

1. Intent Parsing & Style Matching: It first structures the user's brief into five dimensions (topic, content type, audience, tone, title strategy). It then consults its local style library (references/styles.md)—which contains 30+ visual families (e.g., Minimalist Large Type, Photorealistic Texture, 3D Isometric) across 17 content verticals—to recommend 2–3 differentiated design directions with rationale.
2. Text-Free Background Generation: Using the platform's native image generation tool, it creates high-quality, text-free background images (the default canvas is 900×383). The prompt engineering follows structured recipes (references/prompt_recipes.md) and strictly prohibits any letters, numbers, or text to avoid AI typographic errors. It typically produces 2–4 variations for A/B comparison.
3. Precision Title Composition: The Chinese title (plus optional subtitle and label) is cleanly typeset onto the background's designated "title safe area" (default: bottom third) using the compose function of scripts/studio.py. This leverages Pillow and system CJK fonts for crisp, error-free typography—a decisive advantage over prompting AI to "write text in the image."
4. Preview & Delivery: The preview function assembles all generated assets into a self-contained HTML design book, showcasing each style, background, final composed cover, and the prompting recipe. It supports side-by-side A/B testing. Refinements are iterative and focused (e.g., "warm up the color tone"), regenerating only affected variants.

Scope & Considerations

• Platform Dependency: The skill relies on the host platform's built-in image generation and understanding capabilities. In constrained environments, it may gracefully degrade to producing an HTML/CSS-based cover template.
• Local-First Knowledge: All design intelligence—from the style catalog to composition principles (references/design_principles.md)—is bundled within the skill and does not auto-update. It can analyze user-provided reference images but defaults to its internal knowledge base.
• Requires Clear Subject: The workflow is triggered by a concrete topic or reference. It will ask for clarification if the theme is missing rather than inferring one.
• Scope is Covers Only: This tool is specialized for cover design. It does not handle article body layout, analytics, or publishing. When revising, it proposes only one focused change per iteration to avoid major rework.