908.【Generative AI】Creating a “Kansai TV Shopping–Style” Tech Stack Promo Image

Character-Driven Prompt Engineering (Practical Example)

topics: [“GenerativeAI”, “PromptEngineering”, “GitHub”, “GitHubPages”, “Python”, “VSCode”]

📌 Purpose of This Article

In this article, we examine how far generative AI image outputs can be pushed
when cultural context and character traits are specified very strongly.

In article 907, we covered a generic TV advertisement tone.
In 908, we go one step further by explicitly specifying:

• Kansai-style local TV shopping program
• Cheerful “uncle and auntie” characters
• 👍 thumbs-up gesture
• Blazing fire background (intentionally excessive)

The result of these conditions is the image shown at the top.

🎯 Key Points

• The underlying tech stack is exactly the same:
  • GitHub
  • GitHub Pages
  • Python
  • VSCode
• Only the presentation constraints are changed
• Character, gesture, and background instructions strongly affect the final output

Generative AI tends to
prioritize visual and theatrical constraints over technical meaning
when filling the composition.

⚡ Base Prompt Used (English)

A high-energy Japanese TV shopping advertisement, Kansai style.
Cheerful middle-aged Japanese man and woman, both giving a big thumbs-up gesture.
Exaggerated expressions, loud and friendly Kansai vibe.

Background is fully covered with dramatic blazing flames and spark effects.
Top headline text: “GitHub! Python! VSCode!”
Big central punch phrase in Japanese: “ITやで!”
Everything looks over-the-top, flashy, and intentionally excessive.

Include generic icon-like visuals for:
- GitHub
- Python
- A code editor (VSCode-like)

Poster layout, bold outlines, saturated colors.
No real brand logos, use look-alike icons.
High resolution, sharp readable text, no watermark.

🧩 Instructions That Had the Strongest Effect

The following instructions were particularly effective.

1️⃣ Gesture Specification

both giving a big thumbs-up gesture

• Fixes the pose of the characters
• Greatly stabilizes the perceived character identity
• Without this, hand poses often become ambiguous

2️⃣ Concrete Background Instead of Abstract Words

Background is fully covered with dramatic blazing flames

• Concrete nouns like fire, flames, explosions
  are reproduced far more reliably than abstract words like “energetic”

3️⃣ Explicit Cultural Context

Kansai style
loud and friendly Kansai vibe

This affects:

• Japanese text phrasing
• Degree of facial exaggeration
• Color saturation and visual density

⚠ Common Failure Patterns

🧪 Minimal Prompt Version

Kansai-style Japanese TV shopping ad.
Cheerful middle-aged man and woman, both thumbs-up.
Blazing fire background, loud and flashy.
Headline: “GitHub! Python! VSCode!”
Big Japanese punch text: “ITやで!”
Poster layout, no watermark.

✅ Summary

• Generative AI is highly sensitive to cultural, character, and motion cues
• The tech stack does not need to change to create a completely different impression
• “Technology × Advertising Direction” is a practical communication method

Article 907 shows a general-purpose pattern.
Article 908 is a fully pushed, exaggerated example.

Possible next steps include:

• A reduced-text, higher-stability version
• Exporting Kansai-style humor for an international audience

That concludes this article.