Soft launch Dialecto is still in testing, and details on this site may change as we finish it.

Features

Everything below follows from one decision: Dialecto patches your translations and their call sites surgically, and touches nothing else. Here's what that unlocks — in the order it matters.

Domain-driven editing

Your translations are organized by domain — checkout, onboarding, emails — not one endless list. Browse, search, and edit a domain at a time, with the structure carrying the meaning. And it's full structural editing on your own files: reorder, rename, move strings between domains, add and delete — all of it landing as a clean diff, never a rewrite.

Surgical access — never your whole codebase

The scan sends only your translation files and the small usage snippets we need to be precise. A thin slice, never your whole repo. We take exactly what we need and not a stroke more — you keep the rest. That makes Dialecto the only vendor that doesn't ask you to hand over your codebase. EU-sovereign, with a local-model option, and self-hostable if you want it fully in your hands.

A faster way to edit

A real editor built for translating. Move through strings fast, see where each one is used without leaving your seat, and work in bulk when bulk is what's needed — find and replace, assign, approve. Every operation still produces a minimal diff and still goes through review. Fast, with a paper trail.

Validated before it ships

Nothing sloppy gets through. The completeness merge gate fails a pull request on strings it newly leaves untranslated — configurable per repo, with a "Fix in Dialecto" link straight to the offending strings. Placeholder parity, plural forms, glossary terms, and stale-base detection are checked on every change, and a human reviews every change besides. On top of that sits an advisory AI quality check that scores the translations a pull request adds or changes — it informs your review; blocking is your call to opt into.

AI with curated context

Bring your own AI or use ours — Dialecto is model-agnostic by design. What we own is the grounding: your domain, where each string appears, your brand voice (Dialecto drafts it with you from your own material), your audience personas — tone, formality, market — plus glossary and translation memory. The model sees exactly enough, never a firehose. And whatever it drafts is review-gated: autofill is a per-repo choice, and everything still lands through the pull-request flow.

The rest of the arsenal

How it compares

Translation "platforms" own your strings and regenerate your files. The AI-native command-line tools regenerate them too — just faster. We tested their actual published file-loaders on valid gettext files: one collapsed a 104-line file to 23 broken lines and emptied the plural translations; another changed a live translation on a round trip that should have changed nothing. (We keep the receipts.)

Dimension Translation "platforms" AI-native CLIs Dialecto
What it asks for Your strings, uploaded into their store Your repo, with write access A surgical slice: translation files + the usage snippets we need
Where your strings live Their database Your repo — regenerated on every run Your repo, untouched except for the changes you made
What a change looks like A regenerated file A regenerated file, force-pushed A one-line-per-change diff; comments and order preserved
How strings are organized Flat lists, tags if you're lucky Whatever the file had Domains, first-class
Before anything merges Varies Autofill races ahead Merge gate, validation checks, human review
Renames fix your code too No No Yes — call-site rewrite, same pull request