Create AI‑Generated Code‑Style Business Cards with Prompt Engineering

Background

A recent X post showcased a novel "code‑style" business card where a hand holds a code editor window displaying personal information. The concept quickly went viral, inspiring many users to replicate the design using generative AI models.

Prompt Template

Prompt Template:
A close‑up shot of a hand holding a business card designed to look like a {file_format} file open in {code_editor}. The card shows code formatted in {syntax_type} with keys like {key1}, {key2}, {key3}, and {key4}. The window includes typical toolbar icons and a title bar labeled {file_name}, styled exactly like the interface of {code_editor}. Background is slightly blurred, keeping the focus on the card.

Example placeholders:
{file_format}: JSON, XML, YAML
{code_editor}: Notepad++, VS Code
{syntax_type}: JSON, XML
{key1}: "name": "Your Name"
{key2}: "title": "Your Title"
{key3}: "email": "[email protected]"
{key4}: "link": "yourwebsite"
[{key5}: "phone": "Phone Number"]
{file_name}: Business Card.json or Business Card.xml

How to Use the Prompt

Replace the placeholders with the desired file format, editor, and personal details. For example, set {file_format} to JSON, {code_editor} to VS Code, and fill {key1} – {key4} with your name, title, email, and website. The prompt instructs the model to generate an image where the business card appears as a code snippet inside a realistic editor window.

Model Comparisons

Experiments with different generative models showed varying results:

GPT‑4o (4o) : Handles long, complex prompts well but struggles with intricate Chinese characters, often producing garbled text for characters with many strokes.

iDream 3.0 : Better at rendering Chinese characters accurately, though it sometimes misinterprets the editor layout.

Doubao : Similar issues to 4o, with occasional layout confusion.

The differences stem from the underlying architectures: 4o uses an autoregressive approach that excels with long textual context, whereas diffusion‑based image models focus more on visual description.

Practical Tips

Prefer English for personal details in the prompt to avoid character‑encoding errors.

Specify the aspect ratio (e.g., 2:3) to suit mobile displays.

Iterate on the prompt by adjusting placeholders or adding style cues (e.g., "dark background", "blurred desk").

If a model fails to render certain Chinese characters, switch to a model with stronger multilingual support (e.g., iDream).

Conclusion

By leveraging a well‑structured prompt template, users can quickly generate stylish, geek‑centric business cards that mimic code editors. The approach highlights both the creative potential of generative AI and the importance of prompt engineering when targeting specific visual layouts.