Skip to content
Steven Barash

Localizr

AI translation tuned to a place, dialect, and tone, not just a language.

Role
Creator and developer
Status
Active
Technologies
Next.js, TypeScript, Descope, OpenAI-compatible APIs

Where it stands: Available as a web app with authenticated translation APIs.

Limits: Dialect accuracy depends on the selected model. Review translations before use; Legalese is a tone option, not a certified translation.

Why Localizr exists

A language name alone leaves a lot unspecified: where it is spoken, which expressions feel natural, and how formal the text should sound. Localizr makes those choices explicit.

Choose a language, country, dialect, and tone, then translate with your own model provider.

How it works

  • Descope handles sign-in. You connect your own OpenRouter or OpenAI-compatible provider and choose a model.
  • The server validates the destination against a curated catalog and builds a prompt with regional vocabulary, idioms, and tone instructions while asking the model to preserve the original meaning.
  • Translations stream into the interface. Non-Latin output gets a second pass for phonetic transliteration; only completed results enter browser-local history.
  • Text passes through the Localizr server to an external model provider; this is not an offline translation engine.
Back to projects
Start