Table of Contents

1. Introduction: Underestimated Effort and the Cost of 5000 Credits

"If I use AI, making my blog English-compatible will only take 30 minutes."

That's how lightly I started the blog multilingualization project, but reality wasn't so sweet. The final time spent was 3 hours. And the Genspark credits consumed were approximately 5000 credits (equivalent to about 1,800 JPY).

It certainly wasn't "magically completed with one click," but it's true that I was able to complete "DB design changes," "full article translation," and "SEO implementation," which would normally take several days to a week, in half a day.

This article will reveal the failure story of being at the mercy of AI's rampage, and the eventually successful "actual translation script (Python)."

Blog Translation Automation System Workflow

Figure 1: Overall workflow of the blog translation automation system (DB acquisition → Gemini translation → Link replacement → DB storage)

2. [Strategy Phase] Having AI Write "Specifications" and Brainstorming

I don't immediately use Genspark's "Super Agent (paid feature)." That's because credits melt away quickly. First, I thoroughly consulted (brainstormed) with Claude and Gemini on the free chat screen.

Blog Structure Diagram - Japanese and English Version Directory Structure

Figure 2: Blog structure after multilingualization (Japanese version: /post/, English version: /en/post/)

Part of the "Implementation Specifications" Generated by AI

As a result of the brainstorming, Genspark output precise specifications like the following. Basing the implementation on this minimized rework.

## Implementation Strategy Selected method: Single Database + Language Field + Slug Suffix Method ## Database Design (Extended Schema) ALTER TABLE blog_posts ADD COLUMN language TEXT NOT NULL DEFAULT 'ja'; ALTER TABLE blog_posts ADD COLUMN original_post_id INTEGER; ## Important Rules - Japanese article slug: '01-genspark_lies' - English article slug: '01-genspark_lies-en' (-en added to avoid UNIQUE constraint)
Tips: Completing this "design consultation" within the free tier is the biggest secret to saving credits.

3. [Failure] The Rampage of Batch Automatic Translation by Genspark Agent

Once the specifications were set, I enthusiastically instructed the Super Agent:
"Write and execute a script to batch translate all articles and register them in the DB!"

This was a mistake.

The "batch processing script" generated by the AI timed out midway or had forgotten closing tags, resulting in a large number of partially translated "strange articles" being generated in the database.

Lesson from failure: Do not entrust AI with "large-scale batch processing." Errors will inevitably occur, and recovery (e.g., deleting junk data) will take twice as long.

4. [Solution] Explaining the Code for Automatic Blog Post Translation with Python and Gemini API

"Haste makes waste." I switched strategies and had a "Python script that carefully translates and registers a single article" created.

Below is the full text of the script I actually used. How did I execute it? I simply pasted it into Genspark's sandbox (AI Developer environment) and clicked "Run." There's no need to even set up a Python environment on your own computer.

Note: If you use the following code, please rewrite the Cloudflare D1 database name (genspark-blog-db) and table names to match your own environment.
[Click to expand] Full Python script (translate_single_article.py) 
#!/usr/bin/env python3
# Single article translation script
# Usage: python3 translate_single_article.py <slug>

import json, subprocess, sys, re

# Authentication information (management via environment variables recommended)
CLOUDFLARE_EMAIL = "[email protected]"
CLOUDFLARE_API_KEY = "your-api-key"
GEMINI_API_KEY = "your-gemini-api-key"

def translate_with_gemini(text, instruction):
    import google.generativeai as genai
    genai.configure(api_key=GEMINI_API_KEY)
    model = genai.GenerativeModel('gemini-2.5-flash')
    return model.generate_content(f"{instruction}\n\n{text}").text.strip()

def translate_article(article):
    title = translate_with_gemini(article['title'], "Translate to English.")
    
    instruction = """Translate to English.
REQUIREMENTS:
1. Preserve ALL HTML
2. Convert: /post/XXX to /en/post/XXX-en
3. Translate comments, NOT code"""
    
    content = translate_with_gemini(article['content'], instruction)
    content = re.sub(r'/post/([^"\\s]+)(?!-en)', r'/en/post/\\1-en', content)
    
    return {'title': title, 'content': content}

# Main processing omitted
No environment setup required: This script does not require you to set up a Python environment on your own PC. It runs simply by pasting it into Genspark's AI Developer (sandbox) environment and executing it.

This script is very simple, yet powerful.

  • By using Gemini 2.5 Flash, translation costs are almost zero.
  • Automatically rewrite internal links (/post/) to the English version (/en/post/) using regular expressions.
  • By including Alt attributes in the translation target, SEO measures are perfect.

Execute this one article at a time and confirm the display in the browser. This "gritty" procedure was ultimately the most reliable and fastest shortcut.

5. Recommendations for WordPress Users: How to Use Genspark?

I built this blog system itself with Genspark (Cloudflare D1 + Hono), so I could freely change the database structure, but I believe many readers use WordPress.

For WordPress users looking to multilingualize with Genspark, directly manipulating the database is too risky. The two safer and more realistic approaches are:

  1. Markdown Copy-Paste Strategy (Recommended):
    Give Genspark the URL of your Japanese article and instruct it, "Translate this article into an SEO-conscious English article and output it in Markdown format." Copy-pasting the output content into a new WordPress article and linking it with language settings (e.g., Polylang) is the safest approach.
  2. XML Import Strategy (Advanced Users):
    Export articles from WordPress in XML format, ask Genspark to "replace only the body tags within the XML with English," and then re-import them. This allows for batch processing, but a backup is essential.

6. Summary: It's Gritty, But the Door to the World Has Opened

What should have taken 30 minutes ended up taking 3 hours, consuming 5000 credits (approx. 1,800 JPY), and I even experienced a DB reset. It certainly wasn't smart development.

However, if you consider that I obtained a "complete multilingual blog with SEO support" for just 1,800 JPY and 3 hours, the cost-performance is overwhelming.
If I had outsourced or translated manually, it would have taken tens of thousands of yen and several weeks.

Genspark is not a magic wand, but an "exceptionally competent assistant that handles gritty tasks 100 times faster."
Why don't you also invest a little credit and time to knock on the door to the world?

Related Articles