A few weeks ago I ran into a small but frustrating problem.
I was using AI to translate JSON i18n files. It worked… until it didn’t.
Keys would get changed
Placeholders like {name} or %d would break
Sometimes the JSON wouldn’t even parse
Every time, I had to manually fix the output before I could use it.
That defeats the whole point.
The real issue
AI is great at generating text, but it doesn’t respect strict structure.
And i18n files depend on structure:
keys must stay identical
placeholders must remain untouched
formatting must be valid
If any of those break, your app breaks.
What I built
I built a simple tool that does one thing:
Translate JSON i18n files without breaking anything
You upload a file, choose a language, and get back:
same keys
placeholders preserved
valid JSON
production-ready output
No cleanup needed.
Where most tools fail
The interesting part wasn’t building it, it was testing it properly.
I tested against real-world cases:
deeply nested objects
arrays
mixed placeholders ({name}, %d, {count})
UI strings, emails, notifications
This is where most approaches fail quietly.
Why I made it
I didn’t want a smarter translator.
I wanted a safer one.
Something you can actually trust in a production workflow without double-checking every line.
Would you use this?
I’m trying to understand:
Is this something you’d actually use in your workflow?
How are you currently handling JSON translations?
What would make this a must-have tool for you?
Try it
If you want to test it:
Top comments (0)