SummaryPretty Prompt and Globalize
The requirement
Keep a fast-moving web app translated in every language, on every release, with two founders and no localization owner.
The solution
Pretty Prompt connected its GitHub repository to Globalize in May 2026. Every push with new copy comes back translated, as a pull request.
The results
- Mixed-language support tickets went from about a dozen to zero.
- About 70 translation jobs ran from May to September with nobody starting them.
- Translation memory covered 92–100% of strings on most jobs.
AboutPretty Prompt
Pretty Prompt is an AI prompt enhancer for ChatGPT, Claude, Gemini and 15+ other AI providers that turns a rough idea into a clear, structured prompt. It ships as a Chrome extension, a web app and an MCP server, run by co-founders Charlie Day and Ilai Szpiezak from London. It went from 0 to 4,000 users in under a week and passed 50,000 in just over a year.
GrowthInternational growth made translation a priority
Pretty Prompt never targeted a region. Then a Korean YouTube video featured the product, more than 10,000 new users signed up in March 2026, and hundreds of them started paying.
"We suddenly got hundreds of people starting to pay for our product. We had an enormous spike from South Korea. That's where our realization came from that we need to expand to more languages."Charlie Day, co-founder
Users write their idea in whatever language feels natural, and Pretty Prompt improves it in that same language. Some made the language a condition of paying: several asked for Traditional Chinese and said they would buy if the app came in it. Traditional Chinese is live today.
"You don't know which users you would have missed because it wasn't translated."Charlie Day
ChallengeKeeping translations in sync with every new feature
Pretty Prompt already translated its product with AI tools. Running a translation took a minute or two. The hard part was remembering to run it after every copy change, then checking the result in languages nobody on the team speaks.
"Actually running it didn't take long (maybe a minute or two), but the real cost was remembering to do it every time (constantly forgetting) and being confident that it was correct."Charlie Day
When a feature shipped before its translations, users saw two languages on one screen and wrote in.
"That leads to a bad product experience. It's almost better if it's completely in English or completely in German, because half and half looks like you don't have control over the product."
Charlie DayCo-founder, Pretty PromptSolutionLocalization that runs on every release
Pretty Prompt connected its web app's GitHub repository to Globalize on 11 May 2026, and the first translation job ran the same day. During setup, Globalize converted the app's hard-coded copy into internationalization files in a single pull request. From then on, translation became part of the release instead of a step someone has to remember.
"It just feels like something that should be automated. Globalize is kind of like a safety net for localization."Charlie Day
- The team updates the appNew features and copy changes go into the app's English locale file, as normal.
- Globalize runs on the pushIt finds new or changed strings, reuses existing translations, translates the rest and runs an automated quality check.
- Translations come back to GitHubAs a pull request, or a commit on the same feature branch.
- The changes shipThe team reviews and merges, and the update goes out in every language at once.
- de.jsonDeutsch
- ko.json한국어
- zh-TW.json繁體中文
- ja.json日本語
- ar.jsonالعربية
- hi.jsonहिन्दी
- and 22 more locale files
Illustration of a Globalize pull request, built from Pretty Prompt's job data. Not a screenshot.
"For me, the PR loop covers everything."Charlie Day
ResultsThe half-translated screens stopped
Mixed-language support tickets. About a dozen before Globalize, none since.
Strings under management
Pretty Prompt web app, 2026. No localization hire.
| Before | With Globalize | |
|---|---|---|
| Translating a copy change | An AI prompt, run by hand | Starts on every push |
| Who has to remember | One of the two founders | Nobody |
| Checking languages nobody speaks | Trust the output | Automated quality check on every job |
Between May and September 2026, Globalize ran about 70 translation jobs through Pretty Prompt's GitHub workflow while the project grew from 84 strings to 6,966. For Charlie and his team, that means one less thing to manage and confidence that no release goes out half translated.
All of it runs on Globalize's entry-level Indie plan: a 50,000-user app in 29 languages, with no sales call and no custom contract.
FitWho this is for, in Charlie's words
"If all your users are in English, you don't need to translate."
That is where Charlie starts when another small team asks him. For a product like his, with tens of thousands of users arriving from anywhere, language is one of the few things that makes the product better for everyone at once.
"If you've got 50 to 60,000 users, you can't have an account executive for all of those. Adding languages is probably one of the biggest things we can do that is low-touch."Charlie Day
"If you have lots of users from all over the world, using Globalize is great for casting a wide net and ensuring the net is really tight so you don't lose any fish from a localization point of view."Charlie Day