How AI and RPA Supercharged Film‑Style Title Graphics and Cover Design at Scale

Background and Goals

The team observed that top overseas video products rely on striking "artistic title graphics" to reinforce IP recognition and memory. To boost the attractiveness of their own story, music, and drama cards on the Haokan video platform, they introduced a systematic AI‑driven title‑design pipeline.

AI Art Font Design – Three‑Step Methodology

Step 1: Text Pre‑processing

Raw spreadsheet titles are cleaned, keeping only the title column. An AI model automatically classifies titles and generates an initial prompt for each category.

Step 2: AI‑Generated Artistic Fonts

Prompt: 艺术字体设计+标题“xx”，字体描述词（风格\质感\色彩）+黑底白字

Short titles (≤10 characters) work best; longer texts require manual spacing to avoid semantic errors.

Step 3: Reusable Prompt Templates

After multiple iterations, five genre‑specific prompt templates were distilled (modern romance, youth campus, realistic adaptation, ethical family, ancient romance). Each template encodes style, texture, and color cues that can be swapped with a new title to produce consistent fonts.

AI Cover Design – Formula Extraction

Analysis of popular short‑drama covers revealed three high‑impact elements: exaggerated character expression + action, bold conflicting title, and high‑saturation color clash. These “attraction three‑elements” were encoded into two prompt formulas—one “fool‑proof” and one “premium”—that combine character, scene, and font cues.

Prompt: 故事海报+男/女主视角+标题“xxx”+标题位置，人物表情张力十足，大标题突出，冷暖撞色高对比，电影级光影处理。

Templates were further organized by genre (e.g., modern romance, youth campus, ethical suspense, ancient power struggles) and visual tone, enabling one‑click matching of scene and font style.

Automated Production Workflow

To move from semi‑automatic RPA to full productization, the team built a ComfyUI‑based pipeline that chains prompt expansion, text‑to‑image generation, smart matting, and size‑standardization. The workflow runs end‑to‑end: title input → high‑quality cover generation → automatic review → multi‑platform distribution.

# Role preset
你是一位专注于电影级封面视觉叙事的提示词专家，擅长将用户提供的一句话转化为富有镜头感、情绪张力与美学质感的画面描述。

# Output structure
Prompt：电影剧照，情节情绪描述，人物表情张力十足，鲜明柔和通透，电影级光影处理。

The pipeline was deployed to an internal production platform, linking the CMS to distribution channels and turning design creativity into a scalable, platform‑level asset.

Key Takeaways

AI evolves from a mere efficiency tool to a creative co‑partner that amplifies designers' aesthetic judgment, strategic thinking, and systematization. By iterating from manual tuning → template reuse → RPA automation → ComfyUI productization, the team created a repeatable, evolvable AI visual‑production system capable of delivering hundreds of high‑click‑rate covers quickly and reliably.